A torque vector control method for electric intelligent vehicles

Abstract

This invention discloses a torque vector control method for electric intelligent vehicles, comprising the following steps: First, considering the time-varying characteristics of tire model parameters and longitudinal vehicle speed, an uncertain system model oriented towards control is established. Then, based on the uncertain system model, a higher-level motion integrated control algorithm based on robust model predictive control is designed. The robust control problem satisfying design constraints and minimizing the upper bound of the objective function is transformed into a finite-dimensional convex optimization problem within the LMI framework shown below. An offline optimization plus online synthesis method is designed to improve computational efficiency. Finally, considering the driving safety and economy of electric intelligent vehicles, a torque distribution control algorithm based on multi-objective coordinated optimization is designed, using the weighted sum of tire load rate and motor energy loss as the optimization control objective, and dynamically adjusting the weights of different objectives based on a custom correlation function. This method can adapt to various complex operating conditions, maximizing vehicle driving safety while simultaneously improving vehicle driving economy.

Description

Technical Field

[0001] This invention belongs to the field of intelligent vehicle active safety and motion control technology in vehicle dynamics control, specifically relating to a torque vector control method for electric intelligent vehicles. Background Technology

[0002] With the increasing severity of energy shortages, environmental pollution, and traffic safety issues, the electrification and intelligentization of automobiles are developing rapidly, and electric intelligent vehicles have become one of the most popular research topics in the world today. Four-wheel independent drive electric intelligent vehicles can independently decouple and control the torque of each wheel with rapid and accurate response, maximizing the advantages of advanced driver assistance systems and higher levels of autonomous driving technologies. This highly forward-looking chassis configuration has broad application prospects in future transportation, and research on its key motion control technologies has significant theoretical and engineering application value.

[0003] Torque vector control of electric intelligent vehicles refers to how to control the vehicle to travel along a planned path and ensure the vehicle's driving safety. Scholars at home and abroad have conducted a lot of research on this and proposed many advanced control methods such as optimal control, robust control, and sliding mode control. For example, Chinese Patent 11890951B discloses a trajectory tracking and motion control method for intelligent electric vehicles. It uses a model predictive control algorithm to design a yaw stability controller and then distributes torque. The method disclosed in this invention is effective in ensuring vehicle safety, but it fails to consider the uncertainty of vehicle model parameters and motor energy consumption, resulting in poor overall performance under extremely complex working conditions.

[0004] Despite the numerous methods proposed, many unresolved issues remain in torque vector control for electric intelligent vehicles due to the complex mechanisms of vehicle longitudinal and lateral dynamics and the inherent limitations of various control methods. First, the accuracy of the vehicle model used for controller design significantly impacts controller performance. It is necessary to consider the uncertainties in tire model parameters and longitudinal vehicle speed, construct a high-fidelity vehicle model that facilitates controller design, and design robust control laws. Furthermore, considering the typical overdrive characteristics of electric intelligent vehicles, the operating state of the motor and the resulting unnecessary energy losses must be taken into account when designing the torque vector controller. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the torque vector control of electric intelligent vehicles, such as modeling errors, external disturbances, actuator saturation, complex overdrive characteristics, and energy consumption, and to provide a torque vector control method for electric intelligent vehicles.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a torque vector control method for an electric intelligent vehicle, comprising:

[0007] Step 1: Based on the time-varying characteristics of tire model parameters and longitudinal vehicle speed, establish a discrete multicellular uncertain system model with bounded disturbances;

[0008] Step 2: Based on the discrete multicellular uncertain system model with bounded disturbances in Step 1, design an upper-level motion integrated control algorithm based on robust model predictive control, and obtain the expected values ​​of the front wheel steering angle and additional yaw moment of the electric intelligent vehicle.

[0009] Step 3: Based on the expected value of the additional yaw moment in Step 2, design a torque distribution control algorithm based on multi-objective coordinated optimization, and obtain the expected torque of the four wheels of the electric intelligent vehicle.

[0010] Further, step 1 includes:

[0011] Step 11: Obtain the basic parameters of the vehicle to be controlled, including: vehicle mass m, sprung mass m s yaw moment of inertia I z Tilt moment of inertia I x Front wheelbase l f Rear wheelbase l r The distance h from the roll center to the center of mass r Wheelbase l w Total tilt stiffness K φ Total roll damping C φ Based on the aforementioned basic vehicle parameters, the three-degree-of-freedom dynamic model of the vehicle is established as follows:

[0012]

[0013] Where V x V represents the longitudinal dimension of the vehicle. y φ represents the lateral tilt, ω is the roll angle, and φ is the lateral tilt angle. r It is the yaw rate, δ f It's the front wheel steering angle, F xfl and F xfr F represents the longitudinal tire force. yfl ,F yfr ,F yrl and F yrr Representing lateral tire force, the subscripts fl, fr, rl, and rr of the longitudinal and lateral tire force symbols represent the left front, right front, left rear, and right rear wheels, respectively, ΔM z Represents additional yaw moment;

[0014] Step 12: Based on the relative positions of the vehicle's reference trajectory and actual trajectory, establish the vehicle's kinematic model as follows:

[0015]

[0016] Where D P κ is the aiming distance, e is the reference path curvature, and κ is the reference path curvature. y For lateral aiming deviation, β is the heading angle deviation, and β is the centroid sideslip angle;

[0017] Step 13: Define the system state vector Control input vector u = [δ f ΔM z ] T Output vector Interference vector w = [κ] T Based on the three-degree-of-freedom dynamics model of the vehicle in step 11 and the vehicle kinematics model in step 12, the system state-space equations under ideal conditions are constructed as follows:

[0018]

[0019] in

[0020]

[0021] E = [0V] x 0000] T ;

[0022] Step 14: Define the time-varying parameter ο1 = V x , ο4=ρ f ,ο5=ρ r The time-varying parameter varies within the range of: ο1∈[ο 1min ,ο 1max ],ο2∈[ο 2min ,ο 2max ],ο3∈[ο 3min ,ο 3max ],ο4∈[ο 4min ,ο 4max ],ο5∈[ο 5min ,ο 5max [and determine the linear parametric form of matrices A, B, and E:]

[0023]

[0024] Step 15: Define a time-varying parameter vector O = [ο1,ο2,ο3,ο4,ο5], which has eight possible combinations:

[0025] O1 = [ο 1max ,ο 2max ,ο 3max ,ο 4max ,ο 5max ],

[0026] O2 = [ο 1min ,ο 2max ,ο 3max ,ο 4max ,ο 5max ], ...

[0028] O8 = [ο 1min ,ο 2min ,ο 3min ,ο 4min ,ο 5min ].

[0029] Based on the above eight time-varying parameter vector combination forms and the linear parameter-varying forms of matrices A, B, and E described in step 14, the local matrices at the vertices of the multicellular system are determined as follows:

[0030]

[0031] Step 16: Determine the vertex weight coefficients of the multicellular system:

[0032] h1=(1-σ1)(1-σ2)(1-σ3)

[0033] h2=σ1(1-σ2)(1-σ3) ...

[0035] h8=σ1σ2σ3

[0036] in

[0037] Step 17: Based on the local matrices at the vertices of the multicellular system in Step 15 and the weight coefficients of each vertex of the multicellular system in Step 16, establish a model of a disturbed multicellular linear time-varying continuous system:

[0038]

[0039] Step 18: Discretize the continuous system perturbed multicellular linear time-varying system model from Step 17 to obtain a discrete multicellular uncertain system model with bounded perturbations:

[0040]

[0041] y(k+1)=C(k)x(k)

[0042] Where x(k) represents the system state vector at the current time k after discretization, u(k) represents the actual control law of the system, and w(k) represents the disturbance vector at the current time k after discretization.

[0043] This represents the local matrix at the vertex of a discrete multicellular uncertain system model with bounded disturbances.

[0044] Furthermore, step 2, which involves designing an upper-level motion integrated control algorithm based on robust model predictive control, includes:

[0045] Step 21: Optimize the design; construct a finite-dimensional convex optimization problem within the LMI framework;

[0046] Step 22: Offline solution: Solve the finite-dimensional convex optimization problem within the LMI framework and obtain a complete lookup table;

[0047] Step 23: Online Synthesis: Perform a bisecting search in the complete lookup table and solve for the discretized control input vector u(k) at the current time k.

[0048] Furthermore, the specific steps of the optimization design include:

[0049] Step 211: Design the objective function of the upper-level motion integrated control algorithm:

[0050]

[0051] Where Q and R are preset weight matrices, and i is the cumulative index;

[0052] Step 212: Design the constraints of the upper-level motion integrated control algorithm, wherein the constraints of the upper-level motion integrated control algorithm include system state constraints, control input constraints and H∞ performance index constraints;

[0053] The system state constraint expression is as follows:

[0054] |β|≈|V y / V x |≤β max

[0055] |α r |≈|(V y -l r ω r ) / V x |≤α rmax

[0056]

[0057] Where β, α r LTR and β represent the center of gravity sideslip angle, rear wheel sideslip angle, and rollover index, respectively. max α rmax and LTR max The preset maximum values ​​corresponding to the center-of-gravity sideslip angle, rear wheel sideslip angle, and rollover index are defined, and the system state constraints are rewritten in matrix form:

[0058]

[0059] in

[0060]

[0061] The control input constraint expression is as follows:

[0062] |δ f |≤δ f,max ,|ΔM z |≤ΔM z,max ,|ΔM x |≤ΔM x,max

[0063] Where, δ f,max ΔM z,max and ΔM x,max These represent the maximum front wheel steering angle, the maximum additional yaw moment, and the maximum additional roll moment under actuator performance constraints, respectively.

[0064] The performance index constraint expression for H∞ is as follows:

[0065]

[0066] Step 213: Determine the upper bound of the objective function mentioned in step 211: maxJ p (k)≤V(x(k|k))≤λ, where λ is the upper bound of the objective function;

[0067] Step 214: The robust control problem that satisfies the constraints of the upper-level motion integrated control algorithm and minimizes the upper bound of the objective function is transformed into a finite-dimensional convex optimization problem within the LMI framework. The finite-dimensional convex optimization problem within the LMI framework is as follows:

[0068]

[0069] st

[0070]

[0071] Furthermore, the specific steps of the offline solution include:

[0072] Step 221: Determine the longitudinal velocity V x and lateral stiffness adjustment factor ρ f and ρ r The range of variation is determined, and the local matrix at the vertex of the discrete multicellular uncertain system model with bounded disturbance is calculated based on it to obtain the calculated local matrix;

[0073] Step 222: Select a series of system state variables \(x\) i , \(i = 1, 2, \ldots, N\);

[0074] Step 223: For the system state variable \(x\) i , and then solve the finite-dimensional convex optimization problem within the LMI framework through the calculated local matrix in Step 221 to obtain a set of matrices \(X\) i and \(Y\) i , store them and their corresponding index \(i\) in the lookup table;

[0075] Step 224: Calculate the state feedback matrix \(K\) corresponding to the matrices \(X\) i and \(Y\) i in Step 222 i \(= Y\) i \(X\) i -1 , and store \(K\) i and its corresponding index \(i\) in the lookup table as well;

[0076] Step 225: If \(i < N\), then select a state \(x\) that satisfies , and let \(i = i + 1\), then loop and execute Steps 223 to 225. If \(i = N\), then end the loop and execute Steps 223 to 225 to obtain a complete lookup table. i+1

[0077] Furthermore, the specific steps of the online synthesis include:

[0078] Step 231: Obtain the system state variable \(x(k)\) at the current time \(k\) in real time;

[0079] Step 232: Perform a binary search in the complete lookup table to determine the largest index \(i\) that satisfies , and obtain the matrices \(X\) max , \(Y\) max and \(K\) i corresponding to the largest index \(i\) i ; i

[0080] Step 233: If \(i < N\), then solve to obtain the continuous coefficient \(\alpha\) i . If \(i = N\), then let \(\alpha\) i \(= 1\);

[0081] Step 234: Substitute the continuous coefficient \(\alpha\) i obtained in Step 233 into the actual control law of the system ​​Solve for the actual control law u(k) of the system. The actual control law u(k) is actually a two-dimensional column vector, where the first element is the expected value of the front wheel steering angle and the second element is the expected value of the additional yaw moment.

[0082] Furthermore, step 3, which involves designing a torque distribution control algorithm based on multi-objective coordinated optimization, includes:

[0083] Step 31: Considering vehicle stability and motor energy loss, the objective function for the lower-level torque distribution control is designed as follows:

[0084] J = νJ1 + (1-ν)J2

[0085] The objective function J consists of two parts, the first part being... z = [T] xij ] T , r e Where μ is the effective radius of the wheel, F is the tire-road adhesion coefficient, and F is the effective radius of the wheel. zij T is the vertical force on the tire. xij For the desired output torque of the motor, the table below shows ij = fl, fr, rl, rr, where fl, fr, rl, and rr represent the front left, front right, rear left, and rear right wheels, respectively.

[0086] The second part of the objective function J

[0087] w ij For the motor speed, η(T) xij ,w ij P represents the motor efficiency. max This represents the maximum power of the motor.

[0088] In the objective function J, ν is the weighting coefficient that coordinates the two parts J1 and J2;

[0089] Step 32: Determine the constraints of the lower-level torque distribution control, which include control expectation constraints and control output constraints;

[0090] The expression for the control expectation constraint is:

[0091]

[0092] Where F xreq ΔM represents the total expected longitudinal force. z Represents additional yaw moment;

[0093] The control output constraint expression is:

[0094]

[0095] Where T dmax and T bmax These represent the maximum torque for motor drive and braking, respectively.

[0096] Step 33: Based on the objective function described in Step 31 and the constraints of the lower-level torque distribution control described in Step 32, the expression for the lower-level torque distribution control optimization problem is constructed as follows:

[0097]

[0098] in

[0099]

[0100] ub=min|(μr e F zij ,T dmax )|,lb=-min|(μr e F zij ,T bmax )|

[0101] The interior point method is used to solve the lower-level torque distribution control optimization problem to obtain the desired torque of the four wheels.

[0102] Furthermore, the step of determining the weighting coefficients ν of J1 and J2 in step 31 includes:

[0103] Step 311: Define the relevant functions Where a β and b β These are the preset stability domain boundary parameters;

[0104] Step 312: Obtain the current vehicle's sideslip angle β and sideslip rate. And based on that, calculate step 311 in the current state. The value;

[0105] Step 313: Based on the calculation in step 312 The value, calculate

[0106] Among them, ο m This is the preset maximum value of the relevant function.

[0107] Beneficial effects:

[0108] 1. By considering tire model parameters and longitudinal vehicle speed time-varying characteristics, an upper-level motion integrated control algorithm based on robust model predictive control is designed, which effectively improves the driving safety of electric intelligent vehicles. This method can adapt to a variety of complex working conditions and helps reduce potential accident risks.

[0109] 2. Using the weighted sum of tire load rate and motor energy loss as the optimization control objective, combined with a multi-objective coordinated optimization torque distribution control algorithm, the vehicle's driving economy can be effectively managed and optimized. By dynamically adjusting the weights of different objectives, the relationship between safety and economy can be balanced. This method minimizes vehicle energy consumption and improves vehicle driving economy.

[0110] 3. By designing an offline optimization and online integration method, the computational efficiency of the control algorithm was improved, enabling electric intelligent vehicles to react faster and more accurately in actual driving, thereby improving overall performance. Detailed Implementation

[0111] The invention will now be further explained with reference to the accompanying drawings.

[0112] like Figure 1 As shown, the present invention provides a torque vector control method for electric intelligent vehicles, comprising:

[0113] Step 1: Based on the time-varying characteristics of tire model parameters and longitudinal vehicle speed, establish a discrete multicellular uncertain system model with bounded disturbances.

[0114] Step 2: Based on the discrete multicellular uncertain system model with bounded disturbances in Step 1, design an upper-level motion integrated control algorithm based on robust model predictive control, and obtain the expected values ​​of the front wheel steering angle and additional yaw moment of the electric intelligent vehicle.

[0115] Step 3: Based on the expected value of the additional yaw moment in Step 2, design a torque distribution control algorithm based on multi-objective coordinated optimization, and obtain the expected torque of the four wheels of the electric intelligent vehicle.

[0116] like Figure 2 As shown, step 1, establishing a control-oriented uncertain system model includes the following steps:

[0117] Step 11: Establish the vehicle dynamics model.

[0118] Step 12: Establish the vehicle kinematics model.

[0119] Step 13: Construct the state-space equations of the system under ideal conditions.

[0120] Step 14: Determine the linear parametric form of matrices A, B, and E.

[0121] Step 15: Determine the local matrix form at the vertices of the multicellular system.

[0122] Step 16: Determine the weight coefficients of each vertex in the multicellular system.

[0123] Step 17: Establish a model of a continuous system with a disturbed multicellular linear time-varying system.

[0124] Step 18: Establish a discrete multicellular uncertainty system model with bounded disturbances.

[0125] like Figure 3 As shown, in step 11, the basic parameters of the vehicle to be controlled are obtained. The basic parameters of the vehicle include: vehicle mass m, sprung mass m s yaw moment of inertia I z Tilt moment of inertia I x Front and rear wheelbase l f l r The distance h from the roll center to the center of mass r Wheelbase l w Total tilt stiffness K φ Total roll damping C φ Based on the aforementioned basic vehicle parameters, the three-degree-of-freedom dynamic model of the vehicle is established as follows:

[0126]

[0127] Where V x V y These represent the longitudinal and lateral velocities, respectively; φ is the roll angle; and ω... r It is the yaw rate, δ f It's the front wheel steering angle, F xfl and F xfr F represents the longitudinal tire force. yfl ,F yfr ,F yrl and F yrr The longitudinal and lateral tire forces are represented by the subscripts fl, fr, rl, and rr, which represent the left front, right front, left rear, and right rear tires, respectively. The additional yaw moment ΔM in the dynamic model represents the lateral tire force. z The expression is as follows:

[0128]

[0129] The additional yaw moment ΔM z Longitudinal tire force F xfl ,F xfr ,F xrl and F xrr The following relationship exists between the motor torque and the torque:

[0130]

[0131] Where r e T is the effective radius of the wheel. xij The desired output torque of the motor is to be determined.

[0132] The lateral tire forces in the vehicle dynamics model are represented by the following quasi-linear tire model:

[0133] F yfl =F yfr =C f ρ f α f ,F yrl =F yrr =C r ρ r α r

[0134] Where C f and C r ρ represents the standard value of the lateral stiffness of the front and rear wheels. f and ρ r The adjustment factor represents the front and rear wheel lateral stiffness, which characterizes the uncertainty of the tire model parameters. It is time-varying and bounded.

[0135] like Figure 4 As shown, in step 12, based on the relative positions of the vehicle's reference trajectory and the actual trajectory, the vehicle kinematic model is established as follows:

[0136]

[0137] Where D P κ is the aiming distance, e is the reference path curvature, and κ is the reference path curvature. y For lateral aiming deviation, This represents the heading angle deviation.

[0138] In step 13, the system state vector is defined. Control input vector u = [δ f ΔM z ] T Output vector Interference vector w = [κ] T Based on the three-degree-of-freedom dynamics model of the vehicle in step 11 and the vehicle kinematics model in step 12, the system state-space equations under ideal conditions are constructed as follows:

[0139]

[0140] in

[0141]

[0142] E = [0V] x 0000] T .

[0143] In step 14, time-varying parameters are defined. ο4=ρ f ,ο5=ρ r The time-varying parameter varies within the range of: ο1∈[ο 1min ,ο 1max ],ο2∈[ο 2min ,ο 2max ],ο3∈[ο 3min ,ο 3max ],ο4∈[ο 4min ,ο 4max ],ο5∈[ο 5min ,ο 5max [and determine the linear parametric form of matrices A, B, and E:]

[0144]

[0145] In step 15, a time-varying parameter vector O = [ο1,ο2,ο3,ο4,ο5] is defined, and the time-varying parameter vector has eight possible combinations:

[0146] O1 = [ο 1max ,ο 2max ,ο 3max ,ο 4max ,ο 5max ],

[0147] O2 = [ο 1min ,ο 2max ,ο 3max ,ο 4max ,ο 5max ], ...

[0149] O8 = [ο 1min ,ο 2min ,ο 3min ,ο 4min ,ο 5min ].

[0150] Based on the above eight time-varying parameter vector combination forms and the linear parameter-varying forms of matrices A, B, and E described in step 14, the local matrices at the vertices of the multicellular system are determined as follows:

[0151]

[0152] In step 16, the weight coefficients of each vertex in the multicellular system are determined:

[0153] h1=(1-σ1)(1-σ2)(1-σ3)

[0154] h2=σ1(1-σ2)(1-σ3) ...

[0156] h8=σ1σ2σ3

[0157] in

[0158] In step 17, based on the local matrices at the vertices of the multicellular system in step 15 and the weight coefficients of each vertex of the multicellular system in step 16, a model of a disturbed multicellular linear time-varying continuous system is established:

[0159]

[0160] y(t)=Cx(t)

[0161] In step 18, the continuous system model of the disturbed multicellular linear time-varying system in step 17 is discretized to obtain a discrete multicellular uncertain system model with bounded disturbances:

[0162]

[0163] y(k+1)=C(k)x(k)

[0164] Where x(k) represents the system state vector at the current time k after discretization, u(k) represents the actual control law of the system, and w(k) represents the disturbance vector at the current time k after discretization.

[0165] This represents the local matrix at the vertex of a discrete multicellular uncertain system model with bounded disturbances.

[0166] like Figure 5 As shown, step 2, which involves designing the upper-level motion integrated control algorithm based on robust model predictive control, includes the following steps:

[0167] Step 21: Optimize the design; obtain the finite-dimensional convex optimization problem within the LMI framework.

[0168] Step 22: Offline solution: Solve the finite-dimensional convex optimization problem within the LMI framework and obtain a complete lookup table.

[0169] Step 23: Online Synthesis: Perform a bisecting search in the complete lookup table and solve for the discretized control input vector u(k) at the current time k.

[0170] like Figure 6 As shown, the specific steps for optimization design include:

[0171] Step 211: Design the objective function of the upper-level motion integrated control algorithm:

[0172]

[0173] Q and R are preset weight matrices.

[0174] Step 212: Design the constraints of the upper-level motion integrated control algorithm, wherein the constraints of the upper-level motion integrated control algorithm include system state constraints, control input constraints and H∞ performance index constraints;

[0175] The system state constraint expression is as follows:

[0176] |β|≈|V y / V x |≤β max

[0177] |α r |≈|(V y -l r ω r ) / V x |≤α rmax

[0178]

[0179] Where β, α r LTR and β represent the center of gravity sideslip angle, rear wheel sideslip angle, and rollover index, respectively. max α rmax and LTR max The preset maximum values ​​corresponding to the center-of-gravity sideslip angle, rear wheel sideslip angle, and rollover index are defined, and the system state constraints are rewritten in matrix form:

[0180]

[0181] in

[0182]

[0183] The control input constraint expression is as follows:

[0184] |δ f |≤δ f,max ,|ΔM z |≤ΔM z,max ,|ΔM x |≤ΔM x,max

[0185] Where, δ f,max ΔM z,max and ΔM x,maxThese represent the maximum front wheel steering angle, the maximum additional yaw moment, and the maximum additional roll moment under actuator performance constraints, respectively.

[0186] The performance index constraint expression for H∞ is as follows:

[0187]

[0188] Step 213: Determine the upper bound of the objective function mentioned in step 211: maxJ p (k)≤V(x(k|k))≤λ, where λ is the upper bound of the objective function.

[0189] Step 214: The robust control problem that satisfies the constraints of the upper-level motion integrated control algorithm and minimizes the upper bound of the objective function is transformed into a finite-dimensional convex optimization problem within the LMI framework. The finite-dimensional convex optimization problem within the LMI framework is as follows:

[0190]

[0191] st

[0192]

[0193] like Figure 7 As shown, the specific steps for offline solution include:

[0194] Step 221: Determine the longitudinal velocity V x and lateral stiffness adjustment factor ρ f and ρ r The range of variation is determined, and the local matrix at the vertex of the discrete multicellular uncertain system model with bounded disturbance is calculated based on it to obtain the calculated local matrix.

[0195] Step 222: Select a series of system state variables x i Let i = 1, 2, ..., N. Then, using the local matrix calculated in step 221, solve the finite-dimensional convex optimization problem within the LMI framework to obtain a set of matrices X. i and Y i Store it and its corresponding index i in the lookup table.

[0196] Step 223: Calculate matrix X as described in step 222 i Y i The corresponding state feedback matrix K i =Y i X i -1 and K i And its corresponding index i is also stored in the lookup table.

[0197] Step 224: If i < N, select a state x that satisfies and let i = i + 1, then loop and execute Steps 222 to 224. If i = N, end the loop and execute Steps 222 to 224 to obtain a complete look-up table.

[0198] As Figure 8 shown, the online synthesis specific steps include:

[0199] Step 231: Obtain the system state quantity x(k) at the current moment k in real time.

[0200] Step 232: Perform a bisection search in the said complete look-up table to determine the largest index i that satisfies to obtain the matrix X max corresponding to the largest index i max , Y i and K i . i .

[0201] Step 233: If i < N, solve to obtain the continuous coefficient α i . If i = N, let α i = 1.

[0202] Step 234: Substitute the continuous coefficient α i solved in Step 233 into the actual system control law to solve the actual system control law u(k). The actual system control law u(k) is actually a two-dimensional column vector, where the first element is the expected value of the front wheel steering angle and the second element is the expected value of the additional yaw torque.

[0203] As Figure 9 shown, the steps of designing the torque distribution control algorithm based on multi-objective coordinated optimization in Step 3 include:

[0204] Step 31: Considering vehicle stability and motor energy loss, design the objective function of the lower-layer torque distribution control as follows:

[0205] J = νJ1 + (1 - ν)J2

[0206] The objective function J includes two parts, where the first part z = [T xij T , r e is the effective wheel radius, μ is the tire-road adhesion coefficient, F zij is the tire vertical force, T xij ​​​For the desired output torque of the motor, the table below shows ij = fl, fr, rl, rr, where fl, fr, rl, and rr represent the front left, front right, rear left, and rear right wheels, respectively.

[0207] The second part of the objective function J

[0208] w ij For the motor speed, η(T) xij ,w ij P represents the motor efficiency. max This represents the maximum power of the motor.

[0209] In the objective function J, ν is the weighting coefficient that coordinates the two parts J1 and J2.

[0210] Step 32: Determine the constraints of the lower-level torque distribution control based on the expected value of the additional yaw moment period. The constraints of the lower-level torque distribution control include control expectation constraints and control output constraints.

[0211] The expression for the control expectation constraint is:

[0212]

[0213] Where F xreq ΔM represents the total expected longitudinal force. z Represents additional yaw moment;

[0214] The control output constraint expression is:

[0215]

[0216] Where T dmax and T bmax These represent the maximum torque for motor drive and braking, respectively.

[0217] Step 33: Based on the objective function described in Step 31 and the constraints of the lower-level torque distribution control described in Step 32, the expression for the lower-level torque distribution control optimization problem is constructed as follows:

[0218]

[0219] in

[0220]

[0221] ub=min|(μr e F zij ,T dmax )|,lb=-min|(μr e F zij ,Tbmax )|

[0222] The lower-level torque distribution control optimization problem is then solved to obtain the desired torque for the four wheels.

[0223] The steps for determining the weight coefficients ν of J1 and J2 include:

[0224] Step 311: Define the relevant functions Where a β and b β These are the preset stability domain boundary parameters;

[0225] Step 312: Obtain the current vehicle's sideslip angle β and sideslip rate. And based on that, calculate step 311 in the current state. The value;

[0226] Step 313: Based on the calculation in step 312 The value, calculate

[0227] This invention designs an upper-level motion integrated control algorithm based on robust model predictive control by considering tire model parameters and longitudinal vehicle speed time-varying characteristics. This effectively improves the driving safety of electric intelligent vehicles. The method can adapt to various complex working conditions and helps reduce potential accident risks. In the upper-level motion integrated control algorithm based on robust model predictive control, the calculation efficiency of the control algorithm is improved by designing an offline optimization plus online synthesis method. This enables electric intelligent vehicles to react faster and more accurately in actual driving, thus improving overall performance.

[0228] This invention also uses the weighted sum of tire load rate and motor energy loss as the optimization control objective, combined with a multi-objective coordinated optimization torque distribution control algorithm, to achieve effective management and optimization of vehicle driving economy. By dynamically adjusting the weights of different objectives, the relationship between safety and economy can be balanced. This method minimizes vehicle energy consumption and improves vehicle driving economy.

[0229] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A torque vector control method for an electric intelligent vehicle, characterized in that, include: Step 1: Based on the time-varying characteristics of tire model parameters and longitudinal vehicle speed, establish a discrete multicellular uncertain system model with bounded disturbances; Step 2: Based on the discrete multicellular uncertain system model with bounded disturbances in Step 1, design an upper-level motion integrated control algorithm based on robust model predictive control, and obtain the expected values ​​of the front wheel steering angle and additional yaw moment of the electric intelligent vehicle. Step 3: Based on the expected value of the additional yaw moment in Step 2, design a torque distribution control algorithm based on multi-objective coordinated optimization, and obtain the expected torque of the four wheels of the electric intelligent vehicle. Step 1 includes: Step 11: Obtain the basic parameters of the vehicle to be controlled, including: vehicle mass m, sprung mass m s yaw moment of inertia I z Tilt moment of inertia I x Front wheelbase l f Rear wheelbase l r The distance h from the roll center to the center of mass r Wheelbase l w Total tilt stiffness K φ Total roll damping C φ Based on the aforementioned basic vehicle parameters, the three-degree-of-freedom dynamic model of the vehicle is established as follows: Among them, V x V represents the longitudinal dimension of the vehicle. y φ represents the lateral tilt, ω is the roll angle, and φ is the lateral tilt angle. r It is the yaw rate, δ f It's the front wheel steering angle, F xfl and F xfr F represents the longitudinal tire force. yfl ,F yfr ,F yrl and F yrr Representing lateral tire force, the subscripts fl, fr, rl, and rr of the longitudinal and lateral tire force symbols represent the left front, right front, left rear, and right rear wheels, respectively, ΔM z Represents additional yaw moment; Step 12: Based on the relative positions of the vehicle's reference trajectory and actual trajectory, establish the vehicle's kinematic model as follows: Among them, D P κ is the aiming distance, e is the reference path curvature, and κ is the reference path curvature. y For lateral aiming deviation, β is the heading angle deviation, and β is the centroid sideslip angle; Step 13: Define the system state vector Control input vector u = [δ f ΔM z ] T Output vector Interference vector w = [κ] T Based on the three-degree-of-freedom dynamics model of the vehicle in step 11 and the vehicle kinematics model in step 12, the system state-space equations under ideal conditions are constructed as follows: in, E=[0V x 0000] T ; Step 14: Define the time-varying parameter ο1 = V x , ο4=ρ f ,ο5=ρ r The time-varying parameter varies within the range of: ο1∈[ο 1min ,ο 1max ],ο2∈[ο 2min ,ο 2max ],ο3∈[ο 3min ,ο 3max ],ο4∈[ο 4min ,ο 4max ],ο5∈[ο 5min ,ο 5max [and determine the linear parametric form of matrices A, B, and E:] Step 15: Define a time-varying parameter vector O = [ο1,ο2,ο3,ο4,ο5], which has eight possible combinations: O1=[o 1max ,the 2max ,the 3max ,the 4max ,the 5max ], O2=[o 1min ,the 2max ,the 3max ,the 4max ,the 5max ], ... O8=[o 1min ,the 2min ,the 3min ,the 4min ,the 5min ] Based on the above eight time-varying parameter vector combination forms and the linear parameter-varying forms of matrices A, B, and E described in step 14, the local matrices at the vertices of the multicellular system are determined as follows: Step 16: Determine the vertex weight coefficients of the multicellular system: h1=(1-σ1)(1-σ2)(1-σ3) h2=σ1(1-σ2)(1-σ3) ... h8=σ1σ2σ3 in Step 17: Based on the local matrices at the vertices of the multicellular system in Step 15 and the weight coefficients of each vertex of the multicellular system in Step 16, establish a model of a disturbed multicellular linear time-varying continuous system: Step 18: Discretize the continuous system perturbed multicellular linear time-varying system model from Step 17 to obtain a discrete multicellular uncertain system model with bounded perturbations: y(k+1)=C(k)x(k) Where x(k) represents the system state vector at the current time k after discretization, u(k) represents the actual control law of the system, and w(k) represents the disturbance vector at the current time k after discretization. This represents the local matrix at the vertex of a discrete multicellular uncertain system model with bounded disturbances.

2. The torque vector control method for electric intelligent vehicles according to claim 1, characterized in that, The steps in step 2 of designing the upper-level motion integrated control algorithm based on robust model predictive control include: Step 21: Optimize the design; construct a finite-dimensional convex optimization problem within the LMI framework; Step 22: Offline solution: Solve the finite-dimensional convex optimization problem within the LMI framework and obtain a complete lookup table; Step 23: Online Synthesis: Perform a bisecting search in the complete lookup table and solve for the discretized control input vector u(k) at the current time k.

3. The torque vector control method for electric intelligent vehicles according to claim 2, characterized in that, The specific steps of the optimization design include: Step 211: Design the objective function of the upper-level motion integrated control algorithm: Where Q and R are preset weight matrices, and i is the cumulative index; Step 212: Design the constraints of the upper-level motion integrated control algorithm, wherein the constraints of the upper-level motion integrated control algorithm include system state constraints, control input constraints and H∞ performance index constraints; The system state constraint expression is as follows: |β|≈|V y / V x |≤β max |a r |≈|(V y -l r oh r ) / V x |≤α rmax Among them, β, α r LTR and β represent the center of gravity sideslip angle, rear wheel sideslip angle, and rollover index, respectively. max α rmax and LTR max The preset maximum values ​​corresponding to the center-of-gravity sideslip angle, rear wheel sideslip angle, and rollover index are defined, and the system state constraints are rewritten in matrix form: in, The control input constraint expression is as follows: |δ f |≤δ f,max ,|ΔM z |≤ΔM z,max ,|ΔM x |≤ΔM x,max Where, δ f,max ΔM z,max and ΔM x,max These represent the maximum front wheel steering angle, the maximum additional yaw moment, and the maximum additional roll moment under actuator performance constraints, respectively. The performance index constraint expression for H∞ is as follows: Step 213: Determine the upper bound of the objective function mentioned in step 211: maxJ p (k)≤V(x(k|k))≤λ, where λ is the upper bound of the objective function; Step 214: The robust control problem that satisfies the constraints of the upper-level motion integrated control algorithm and minimizes the upper bound of the objective function is transformed into a finite-dimensional convex optimization problem within the LMI framework. The finite-dimensional convex optimization problem within the LMI framework is as follows: st 4. The torque vector control method for electric intelligent vehicles according to claim 3, characterized in that, The specific steps for offline solution include: Step 221: Determine the longitudinal velocity V x and lateral stiffness adjustment factor ρ f and ρ r The range of variation is determined, and the local matrix at the vertex of the discrete multicellular uncertain system model with bounded disturbance is calculated based on it to obtain the calculated local matrix; Step 222: Select a series of system state variables x i , i = 1, 2, ..., N; Step 223: For the system state variable x i Then, by using the local matrix calculated in step 221, the finite-dimensional convex optimization problem within the LMI framework is solved to obtain a set of matrices X. i and Y i Store it and its corresponding index i in the lookup table; Step 224: Calculate matrix X as described in step 222 i Y i The corresponding state feedback matrix K i =Y i X i -1 and K i And its corresponding index i is also stored in the lookup table; Step 225: If i < N, then select a state x that satisfies , and set i = i + 1, then loop and execute Steps 223 to 225 again. If i = N, then end the loop and execute Steps 223 to 225 to obtain a complete lookup table.​​ 5. The torque vector control method for electric intelligent vehicles according to claim 4, characterized in that, The specific steps of the online integration include: Step 231: Obtain the system state variable x(k) at the current time k in real time; Step 232: Perform a bisection search on the complete lookup table to determine if the conditions are met. The largest index i max To obtain the maximum index i max The corresponding matrix X i Y i and K i ; Step 233: If i < N, then solve to obtain the continuous coefficient α i , if i = N, then let α i = 1; Step 234: Apply the continuity coefficients α obtained in step 233 i Substitute into the actual control law of the system Solve for the actual control law u(k) of the system. The actual control law u(k) is actually a two-dimensional column vector, where the first element is the expected value of the front wheel steering angle and the second element is the expected value of the additional yaw moment.

6. The torque vector control method for electric intelligent vehicles according to claim 1, characterized in that, The steps in step 3 of designing the torque distribution control algorithm based on multi-objective coordinated optimization include: Step 31: Considering vehicle stability and motor energy loss, the objective function for the lower-level torque distribution control is designed as follows: J = νJ1 + (1-ν)J2 The objective function J consists of two parts, the first part being... z = [T] xij ] T , r e Where μ is the effective radius of the wheel, F is the tire-road adhesion coefficient, and F is the effective radius of the wheel. zij T is the vertical force on the tire. xij For the desired output torque of the motor, the table below shows ij = fl, fr, rl, rr, where fl, fr, rl, and rr represent the front left, front right, rear left, and rear right wheels, respectively. The second part of the objective function J w ij For the motor speed, η(T) xij ,w ij P represents the motor efficiency. max This represents the maximum power of the motor. In the objective function J, ν is the weighting coefficient that coordinates the two parts J1 and J2; Step 32: Determine the constraints of the lower-level torque distribution control, which include control expectation constraints and control output constraints; The expression for the control expectation constraint is: Among them, F xreq ΔM represents the total expected longitudinal force. z Represents additional yaw moment; The control output constraint expression is: Among them, T dmax and T bmax These represent the maximum torque for motor drive and braking, respectively. Step 33: Based on the objective function described in Step 31 and the constraints of the lower-level torque distribution control described in Step 32, the expression for the lower-level torque distribution control optimization problem is constructed as follows: in, ub=min|(μr e F zij ,T dmax )|,lb=-min|(μr e F zij ,T bmax )| The interior point method is used to solve the lower-level torque distribution control optimization problem to obtain the desired torque of the four wheels.

7. The torque vector control method for electric intelligent vehicles according to claim 6, characterized in that, The step of determining the weight coefficients ν of J1 and J2 in step 31 includes: Step 311: Define the relevant functions Where a β and b β These are the preset stability domain boundary parameters; Step 312: Obtain the current vehicle's sideslip angle β and sideslip rate. And based on that, calculate step 311 in the current state. The value; Step 313: Based on the calculation in step 312 The value, calculate Among them, ο m This is the preset maximum value of the relevant function.

Images