Driving path tracking and stability control method for distributed driving electric vehicle

Abstract

The invention relates to the technical field of integrated collaborative optimization control, in particular to a driving path tracking and stability control method for a distributed driving electric vehicle, which comprises the following steps of: constructing a vehicle integrated dynamic model, fusing data driving transverse dynamics and other models, and carrying out dynamic control on the vehicle integrated dynamic model; front wheel steering and four-wheel driving coordination control of the distributed driving electric automobile is realized; discretizing the model to obtain a system state, output, control input and disturbance under discrete time; constructing a model prediction controller multi-objective optimization function containing a system output and control input weighting matrix; the speed tracking weight in the system output weighting matrix is adaptively adjusted according to the vehicle running state, and the longitudinal force distribution weight in the input weighting matrix is controlled according to the wheel slip rate and the like; applying a control input constraint containing a first moment rate limit and a system output constraint introducing a slack variable; and the optimization function and the constraint are converted into a quadratic programming problem with a slack term to be solved, and a front wheel rotation angle and longitudinal traction control instruction of each driving wheel are obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of integrated cooperative optimization control, and more particularly to a driving path tracking and stability control method for a distributed drive electric vehicle. BACKGROUND

[0002] Distributed drive electric vehicles (DDEVs) can effectively enhance the lateral stability and handling safety of an automatic driving system under complex working conditions due to the independent control capability of the driving force of each wheel. By coordinating the torque distribution of each driving wheel, DDEVs can generate an active lateral force moment as a redundant control input, effectively enhancing the stability and dynamic response capability of the vehicle during steering. At present, the control system of DDEVs mostly adopts a hierarchical control structure: the upper controller is responsible for calculating the target steering angle and the generalized control force (such as the lateral force moment and the longitudinal traction force), and the lower controller reasonably distributes the control quantity to the four driving wheels according to the upper layer command. This structure can achieve good control effect under most working conditions, but in the high-speed emergency obstacle avoidance scene or sudden working condition, the communication delay and response lag between layers may lead to a decrease in control accuracy, thereby affecting the stability and safety of the vehicle. At the same time, due to the different objective functions between the upper and lower layers, control conflicts may occur under certain boundary conditions, for example, the upper layer expects to generate a large lateral force moment, while the lower layer is limited by the motor capability or adhesion condition and cannot be realized, thereby reducing the overall control performance of the system. Therefore, it is of great significance to develop an integrated control strategy for distributed drive vehicles.

[0003] Therefore, how to provide a driving path tracking and stability control method for a distributed drive electric vehicle is a problem that those skilled in the art need to solve. SUMMARY

[0004] Therefore, the present application provides a driving path tracking and stability control method for a distributed drive electric vehicle, which aims to solve the above technical problems.

[0005] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0006] A driving path tracking and stability control method for a distributed drive electric vehicle, comprising the following steps:

[0007] An integrated vehicle dynamic model is constructed, which integrates a data-driven lateral dynamics model, a lateral error tracking model, a longitudinal speed tracking model, and a nonlinear tire force estimation model based on a Gaussian process, for realizing coordinated control of front wheel steering and four-wheel drive of the distributed drive electric vehicle;

[0008] discretize the vehicle integrated dynamic model to obtain system states, outputs, control inputs and disturbances in discrete time;

[0009] construct a multi-objective optimization function of the model predictive controller, the multi-objective optimization function comprising a weighted matrix of system outputs and a weighted matrix of control inputs;

[0010] adaptively adjust a speed tracking weight in the weighted matrix of system outputs according to a vehicle operating state;

[0011] adaptively adjust a longitudinal force distribution weight in the weighted matrix of control inputs according to a wheel slip ratio, a wheel vertical load and a road adhesion coefficient;

[0012] apply control input constraints including a speed limit at a first control time and system output constraints introducing a slack variable; and

[0013] transform the multi-objective optimization function and the constraints into a quadratic programming problem with a slack term to obtain control commands including front wheel steering angles and longitudinal tractive forces of each drive wheel.

[0014] Further, the nonlinear tire force estimation model based on the Gaussian process constructs a nonlinear residual, constructs a training data set by taking the front and rear wheel side slip angles as inputs and the nonlinear residual as outputs, and performs posterior inference on the Gaussian process based on the training data set to obtain an estimated value of the nonlinear tire force, wherein the nonlinear residual is the actual collected tire lateral force data minus the estimated value of the linear model.

[0015] Further, the vehicle integrated dynamic model takes a vehicle lateral speed, a yaw angular velocity, a lateral deviation, a heading deviation and a longitudinal speed error as a state vector, and takes front wheel steering angles and longitudinal tractive forces of four drive wheels as a control input vector; wherein the data-driven lateral dynamics model represents the yaw characteristics of the vehicle, the lateral error tracking model reflects the deviation dynamics of the vehicle relative to a reference path, and the longitudinal speed tracking model reflects the error dynamics of the vehicle longitudinal speed.

[0016] Further, the adaptive adjustment of the speed tracking weight in the weighted matrix of system outputs according to the vehicle operating state specifically comprises:

[0017] An adaptive adjustment mechanism of the speed tracking weight is designed based on a hyperbolic tangent function, and the speed tracking weight is dynamically adjusted according to threshold values of a vehicle lateral speed, a yaw angular velocity, a lateral error and a heading error, and adjustment parameters of an adaptive function.

[0018] Further, the longitudinal force distribution weight in the control input weight matrix is adaptively adjusted according to the wheel slip ratio, the wheel vertical load and the road adhesion coefficient, and specifically comprises:

[0019] The slip ratio of the i th tire is defined as a function of the ratio of the vehicle longitudinal speed to the product of the corresponding wheel speed and the tire radius;

[0020] According to the size of the slip ratio, the longitudinal control force weight of the wheel is defined as an exponential function related to the slip ratio, wherein the weight maintains a basic value when the slip ratio is small, and increases exponentially with the slip ratio when the slip ratio exceeds a preset threshold value;

[0021] The longitudinal force distribution weight is also calculated based on the road adhesion coefficient and the vertical load of the i th wheel to obtain an initial weight, which comprehensively considers the vertical load and the road adhesion condition.

[0022] Further, the control input constraint specifically comprises:

[0023] The rate limit is introduced to the control input of the first control time to suppress the mutation of the control amount, and the upper and lower limit ranges are set to the control input of all control times.

[0024] Further, the system output constraint specifically comprises:

[0025] The upper and lower limits are set to the system output, and a relaxation variable is introduced to soften the constraint boundary and improve the feasibility of the optimization problem, wherein the relaxation variable is a one-dimensional unit column vector.

[0026] Further, the conversion of the multi-objective optimization function and the constraint condition into a quadratic programming problem with a relaxation term for solving specifically comprises:

[0027] An optimization variable combination vector is constructed, which includes the control input and the relaxation variable;

[0028] The objective function with the relaxation term is converted into a standard quadratic programming form, which is a minimization objective function including a weight matrix H of the quadratic term and a coefficient vector f of the linear term;

[0029] The upper and lower bound constraints are applied to the optimization variables; and

[0030] The linear and equality constraints are applied to the optimization variables, and the linear constraints are determined by a constraint matrix and a constraint boundary vector.

[0031] Further, in the data-driven lateral dynamics model, a generalized yaw moment is introduced, which is generated by the difference between the driving torques of the left and right wheels, and its calculation formula is related to the longitudinal force components of the front left, front right, rear left and rear right tires.

[0032] Further, the obtained longitudinal traction force of each drive wheel is converted into the output torque of the corresponding drive motor through a predetermined conversion relationship, realizing the decoupling transition between the optimization quantity and the actual execution quantity.

[0033] Compared with the prior art, the present application can be based on the dynamics model of the data-driven distributed drive vehicle. The non-linear part of the tire lateral force is learned by the Gaussian regression process method, which significantly improves the accuracy and generalization ability of the vehicle dynamics model. The distributed vehicle system integration model based on the fusion of data-driven lateral dynamics and lateral and longitudinal error modeling. Multi-objective integrated control of vehicle lateral stability, lateral path tracking and longitudinal speed tracking is realized. The model predictive controller based on the fusion of adaptive weight mechanism is designed. The weight can be dynamically adjusted to adapt to different working conditions, realizing accurate path tracking and stability control of the vehicle in high-speed, extreme and variable driving environments. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only a part of the embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creative labor.

[0035] Figure 1 The flowchart of the driving path tracking and stability control method of the distributed drive electric vehicle of the present application.

[0036] Figure 2 The vehicle data diagram of the present application.

[0037] Figure 3 The lateral error tracking model diagram of the present application. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0039] The application provides a high-speed driving path tracking and stable control method for a distributed drive electric vehicle, which specifically comprises three steps: step one, a dynamic model of the distributed drive vehicle based on data driving; step two, an integrated tracking model of the distributed vehicle based on fusion of data-driven lateral dynamics and error modeling; and step three, design of a model predictive controller with fusion of an adaptive weight mechanism. Figure 1 As shown in the description is as follows:

[0040] Step one: a dynamic model of the distributed drive vehicle based on data driving.

[0041] S1: vehicle dynamics model establishment. According to Newton's law, the general expression can be represented as:

[0042]

[0043] wherein m represents the vehicle mass, I z is the moment of inertia of the vehicle around the vertical axis, v x , v y and w z represent the longitudinal speed, lateral speed and yaw rate of the vehicle respectively, F yf and F yr are the lateral tire forces of the front and rear wheels respectively, M z is the additional yaw moment, l f and l r represent the distances from the vehicle mass center to the front and rear axles respectively, and the vehicle parameter schematic diagram is shown in Figure 2 .

[0044] The experimental results show that the lateral tire force is a nonlinear function of the tire side slip angle, but it can be approximately linear in the small side slip angle range. Therefore, it can be represented as a combination model of the linear part and the unknown nonlinear part:

[0045]

[0046] wherein C f and C r are the side slip stiffnesses of the front and rear wheels respectively, f yf (α f ) and f yr (α r ) are unknown nonlinear tire force parts of the front and rear wheels respectively, and depend on the side slip angles α f and α r of the front and rear wheels respectively.

[0047] The side slip angles of the front and rear wheels can be approximately represented as:

[0048]

[0049] S2: Model reconstruction. Substitute equation (2) and equation (3) into equation (1), which can be re-expressed as follows:

[0050]

[0051] M z is a generalized moment generated by the difference between the driving moments of the left and right wheels. This generalized moment M z is mainly used to realize active yaw moment control of the vehicle, thereby improving its lateral stability and path tracking performance. Its calculation formula is as follows:

[0052]

[0053] where W represents the wheel track of the vehicle (i.e., the lateral distance between the left and right wheels); F xlf ,F xrf ,F xlr ,F xrr represent the longitudinal force components of the left front, right front, left rear, and right rear tires in the vehicle coordinate system, respectively.

[0054] S3: Nonlinear tire force modeling based on Gaussian regression process.

[0055] For the nonlinear lateral forces f yf (α f ) and f yr (α r ), this paper proposes a data-driven modeling method based on Gaussian Process (GP), and the modeling steps are as follows:

[0056] 1. Nonlinear residual construction. Based on the actual collected tire lateral force data, subtract the linear model estimate value to obtain the residual value, defined as follows:

[0057]

[0058] where f yi (α i ) represents the nonlinear residual term of the tire lateral force, is the true tire lateral force.

[0059] 2. Training data set construction. Take the front and rear wheel sideslip angles as input and the residual as output to construct the training set Φ:

[0060]

[0061] where α f , α r , f yf (α f ), f yr (αr ) is a one-dimensional column vector containing N samples.

[0062] 3. Gaussian process prior modeling. GP is used to model the nonlinear function, the expression is as follows:

[0063]

[0064] where d GP (·) ~ GP(m(ξ), K(ξ, ξ') is a Gaussian process with mean m(ξ) and covariance K(ξ, ξ'), and the vector represents the estimated value of the front and rear wheel nonlinear lateral force; the input variable ξ = [α f , α r ] represents the side slip angle of the front and rear wheels, which can be calculated by formula (3); w GP represents the measurement noise in the Gaussian process.

[0065] 4. Gaussian process posterior inference. After observing the training data set Φ, the posterior distribution of the estimated function at the test point ξ is still Gaussian, which is in the form of:

[0066]

[0067] The posterior mean and variance are:

[0068]

[0069] where K(·, ·) is the kernel function matrix, is the observation noise variance, and m(ξ) is the mean function, which is usually set to zero, i.e. m(ξ) = 0.

[0070] The kernel function is selected as the squared exponential kernel (Squared Exponential Kernel), which is defined as follows:

[0071]

[0072] where the hyperparameter set contains: length scale matrix M, signal variance and noise variance can be learned by maximum likelihood estimation.

[0073] Step two: distributed vehicle integrated tracking model based on fusion data driven lateral dynamics and error modeling.

[0074] Unlike traditional hierarchical control strategies, this paper proposes an integrated control structure to achieve coordinated control between front-wheel steering and four-wheel drive in distributed-drive electric vehicles, balancing path tracking accuracy and lateral stability. This structure is based on a high-precision dynamic model that integrates lateral and longitudinal characteristics, serving the controller design.

[0075] Specifically, a data-driven lateral dynamics model can effectively characterize vehicle yaw characteristics and achieve stable control through joint adjustment of the vehicle's lateral velocity and yaw rate. Simultaneously, error modeling is introduced to reflect the vehicle's deviation dynamics relative to the reference path, providing a modeling basis for path tracking control.

[0076] S1: Establishment of the lateral error tracking model.

[0077] Vehicle lateral deviation e y deviation from heading For the key state variables, the error dynamic model is constructed as follows:

[0078]

[0079] in, and These represent the rates of change of lateral error and yaw error, respectively. zd Let ρ be the desired yaw rate and ρ be the road curvature. A schematic diagram is shown below. Figure 3 As shown.

[0080] S2: Establishment of the longitudinal velocity tracking model. To avoid the model complexity and solution burden caused by strong coupling between the longitudinal and transverse directions, this paper adopts the following simplified dynamic model of longitudinal error:

[0081]

[0082] in, and V represents the rate of change of longitudinal velocity error and the longitudinal velocity error, respectively. xd and a xd F represents the reference velocity and reference acceleration, respectively. xt The total longitudinal traction force is approximately expressed as follows:

[0083] F xt =F xlf +F xrf +F xlr +F xrr (14)

[0084] S3: Construction of the integrated system model. Combining the data-driven lateral dynamics model (4), the nonlinear tire force expression (9) modeled by Gaussian process, and the error dynamic model (12)-(14), the unified state-space equations of the system are established as follows:

[0085]

[0086] Where the state vector Includes vehicle lateral velocity, yaw rate, lateral deviation, heading deviation, and longitudinal velocity error; the control input vector is u = [δ]. f ,F xlf ,F xrf ,F xrl ,F xrr ] T This includes the front wheel steering angle and the longitudinal traction force of the four-wheel drive wheels; the disturbance vector is... in and These are the nonlinear lateral forces estimated by GP for the front and rear wheels, respectively, where ρ is the lane curvature and a is the lateral force. xd The desired longitudinal acceleration. Matrix A c B c C c and D c The system's state transition matrix, control input matrix, external disturbance input matrix, and output observation matrix are respectively represented as follows:

[0087]

[0088] Step 3: Design of a model prediction controller incorporating an adaptive weighting mechanism.

[0089] S1: System model discretization. To achieve model predictive control, the continuous-time state-space equation (15) is first discretized using the zero-order hold (ZOH) method, and the sampling period is set to T. s The system can be represented in the following discrete form:

[0090]

[0091] Where x(k), y(k), u k Let w(k) and w(k) represent the system state, output, control input, and disturbance in discrete time, respectively, while A d B d C d and D d The state transition matrix, control input matrix, disturbance input matrix, and output observation matrix, respectively, obtained by discretization using the ZOH method, are represented as follows:

[0092]

[0093] S2: Derivation of the model's prediction time-domain representation. Given the initial state x(k) and the prediction time domain N... p Control time domain N c Under these conditions, the evolution of state variables in the prediction time domain can be recursively expressed as:

[0094]

[0095] Where, N p To predict the time-domain step size, N c To control the step size in the time domain.

[0096] From equation (17), the system output sequence in the prediction time domain can be calculated using the following formula:

[0097] Y = C P X+D P U+E P W (18)

[0098] Among them Y=y(k),y(k+1)…,y(k+N p )] T The output prediction vector is X = x(k), x(k+1)..., x(k+N). p )] T The state prediction vector; U = [u(k), u(k+1)..., u(k+N)] p )] T To control the input vector; W = [w(k), w(k+1)..., w(k+N)] p )] T Let C be the perturbation vector. P D P and E P These are the block matrices of the system, specifically represented as follows:

[0099]

[0100] S3: Establishment of the multi-objective optimization function for the controller. The objective function of the model predictive controller is defined as follows:

[0101] min J=(YY des ) T Q(YY des )+U T RU (19)

[0102] Where J is the objective function, Q is the weighting matrix of the system output, R is the weighting matrix of the control input, and Y... des For the desired system output, these matrices are defined as follows:

[0103]

[0104] The weighting matrix Q in the system output is... i middle, Weights representing lateral velocity; The weight representing the yaw rate, The weights representing the lateral error, The weights representing the heading error, The weights represent the longitudinal velocity error; while in the weighting matrix of the control quantity, It is the weight of the steering angle. (ij = fl, fr, rl, rr) represents the weight of the longitudinal traction force of each drive wheel; in the desired system output matrix Y desi In the middle, v ydes w zdes These represent reference values ​​for lateral velocity and lateral angular velocity, respectively, which are used to ensure the yaw stability of autonomous vehicles. Typically, this can be determined by the steady-state response of the lateral dynamics model (i.e., lateral acceleration is zero) as shown in equation (4). The yaw acceleration is zero. The yaw moment is zero M z =0), and ignoring the learning part, the calculation yields:

[0105]

[0106] in, L is the steady-state steering characteristic coefficient of the vehicle, where L = l f +l r This refers to the vehicle's wheelbase length.

[0107] S4: Design of an adaptive adjustment mechanism for weights in multi-objective optimization.

[0108] To improve vehicle control performance under various operating conditions, especially some extreme conditions, an adaptive mechanism for a weighted system is proposed. This mechanism dynamically adjusts the longitudinal velocity tracking weight in the system output and the tire longitudinal force distribution weight in the control input.

[0109] 1. Adaptive Velocity Tracking Weight Mechanism. For the weighting matrix output by the system, a velocity tracking weight mechanism based on the hyperbolic tangent function (tanh) is designed. The adaptive adjustment mechanism is expressed as follows:

[0110]

[0111] Where v yth ,w zth ,e yth ,eφth These are the threshold values ​​set for lateral velocity, yaw rate, lateral error, and heading error, respectively. Indicates the initial weights for velocity tracking; a s ,b s ,c s ,k s These are the adjustment parameters for the adaptive function.

[0112] This mechanism can dynamically reduce the priority of speed tracking in control optimization based on the operating status, enhance the dominant role of lateral stability control in extreme scenarios, and thus improve the robustness and safety of the overall control system.

[0113] 2. Adaptive Mechanism for Longitudinal Force Distribution Weights. To avoid performance degradation under low adhesion or tire slippage conditions, this paper further designs an adaptive weight adjustment mechanism for longitudinal traction. First, the slip ratio of the ij-th tire is defined as:

[0114]

[0115] Among them, v x Let w be the longitudinal velocity of the vehicle body. ij r corresponds to the rotational speed of the wheel. w This is the tire radius.

[0116] Based on the slip ratio, the longitudinal control force weight of this wheel is defined as follows:

[0117]

[0118] Where, k w These are the adjustment parameters of the weighted adaptive mechanism; This represents the initial weight of the longitudinal force on each wheel, taking into account both the vertical load on the wheel and the road adhesion conditions. Its expression is:

[0119]

[0120] Where μ ij F is the road adhesion coefficient of the ij-th wheel; zij The vertical load on the j-th wheel; k f This is the proportional coefficient for weight adjustment, used to adjust the magnitude of the overall longitudinal force weighting.

[0121] This adaptive weighting mechanism maintains a basic weight when the wheel slip ratio is low; when the slip ratio exceeds a set threshold of 0.1, the weight increases exponentially with the slip ratio, thereby rapidly weakening the longitudinal control force of that wheel. By introducing vertical load and adhesion coefficient information, a more reasonable force distribution can be achieved, effectively suppressing slip and improving the controllability and driving stability of autonomous vehicles under various operating conditions.

[0122] S5: System output and control quantity constraints.

[0123] To ensure the safety of the proposed control strategy in practical applications and the feasibility of the solution, this paper sets the following constraints on the control input and system output:

[0124] 1. Constraints on control inputs. Control inputs are limited to the following upper and lower limits:

[0125] U min ≤U≤U max (25)

[0126] Among them, U min and U max These represent the lower and upper limits of the control output, respectively, and are defined as follows:

[0127]

[0128] The first control time (i=1) is not only subject to conventional upper and lower limits, but also to a rate limit to suppress abrupt changes in the control quantity, defined as follows:

[0129]

[0130] Where u p The control quantity Δu is from the previous sampling period. max The maximum permissible change in the control input within each sampling period; the control quantity at other times (i = 2, ..., N) c Satisfies conventional constraints:

[0131] 2. System Output Constraints. To ensure driving safety, this paper also introduces the following constraint form for the system output, and introduces a slack variable ε to avoid the optimization problem becoming unsolvable due to constraint infeasibility:

[0132]

[0133] in, For dimension N P A unit column vector of size N = 1, with dimensions and prediction step size N. P Consistent. ε is a slack variable used to soften the constraint boundary and improve the feasibility of the optimization problem.

[0134] S6: Quadratic programming solution for the control problem.

[0135] By combining the aforementioned multi-objective optimization problem with constraints, the following objective function with relaxation terms can be constructed:

[0136]

[0137] Where ρ is the weighting matrix of the penalty relaxation variables.

[0138] Furthermore, this optimization problem can be transformed into a standard quadratic programming form as follows:

[0139]

[0140] in, The optimization variable is a combined vector containing the control input U and the slack variable ε; G k H is the weight matrix for the quadratic terms; k The coefficient vector of the linear terms; and To optimize the upper and lower bounds of variables; A cons B is the constraint matrix; cons The boundary vectors are constraints; the definitions of each variable and matrix are as follows:

[0141]

[0142] It is worth noting that in the proposed integrated control architecture, to avoid coupling between the front wheel drive torque and its steering angle, the controller uses the longitudinal force of each wheel as the control output, directly applying it to constraint and optimization solutions. This paper uses the longitudinal traction force at the wheel center of each wheel as the control variable, thus achieving decoupling between the front wheel steering angle and the motor control command. While this approach offers good modeling convenience in optimization algorithms, at the physical execution level, the vehicle actually controls the output torque of each drive motor, making it an indirect control command. To achieve the transition from the optimized quantity (wheel longitudinal force) to the actual executed quantity (motor torque), this paper introduces the following transformation relationship:

[0143]

[0144] The distributed drive vehicle integrated optimization control strategy for high-speed driving stability proposed in this application has the following advantages compared with traditional methods:

[0145] (1) Data-driven distributed driving vehicle dynamics model. By learning the nonlinear part of the tire lateral force through the Gaussian regression process, the accuracy and generalization ability of the vehicle dynamics model are significantly improved.

[0146] (2) A distributed vehicle system integrated model based on fused data-driven lateral dynamics and lateral and longitudinal error modeling. Multi-objective integrated control of vehicle lateral stability, lateral path tracking and longitudinal speed tracking is realized.

[0147] (3) Design of a model predictive controller based on a fusion adaptive weighting mechanism. It can dynamically adjust the weights to adapt to different operating conditions, and realize accurate path tracking and stability control of the vehicle in high-speed, extreme and variable driving environments.

[0148] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0149] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for path tracking and stability control of a distributed drive electric vehicle, characterized in that, Includes the following steps: An integrated vehicle dynamic model is constructed, which integrates a data-driven lateral dynamics model, a lateral error tracking model, a longitudinal speed tracking model, and a nonlinear tire force estimation model based on Gaussian processes, to achieve coordinated control of front-wheel steering and four-wheel drive in a distributed drive electric vehicle. The vehicle integrated dynamic model is discretized to obtain the system state, output, control input, and disturbance in discrete time. Construct a multi-objective optimization function for the model predictive controller, wherein the multi-objective optimization function includes a weighted matrix of the system output and a weighted matrix of the control input; The speed tracking weights in the system output weighting matrix are adaptively adjusted according to the vehicle's operating status. The longitudinal force distribution weights in the control input weighting matrix are adaptively adjusted based on wheel slip ratio, wheel vertical load, and road adhesion coefficient. Apply control input constraints and system output constraints, wherein the control input constraints include a rate limit for the first control time, and the system output constraints introduce slack variables; as well as The multi-objective optimization function and the constraints are transformed into a quadratic programming problem with relaxation terms and solved to obtain control commands including the front wheel steering angle and the longitudinal traction force of each drive wheel.

2. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The nonlinear tire force estimation model based on Gaussian process constructs a training dataset by using the front and rear wheel slip angles as inputs and the nonlinear residuals as outputs. Based on the training dataset, a posterior inference of the Gaussian process is performed to obtain the estimated value of the nonlinear tire force. The nonlinear residual is the actual collected tire lateral force data minus the estimated value of the linear model.

3. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The vehicle integrated dynamic model uses the vehicle's lateral velocity, yaw rate, lateral deviation, heading deviation, and longitudinal velocity error as state vectors, and the front wheel steering angle and the longitudinal traction force of the four-wheel drive wheels as control input vectors. Among them, the data-driven lateral dynamics model characterizes the vehicle's yaw characteristics, the lateral error tracking model reflects the vehicle's deviation dynamics relative to the reference path, and the longitudinal velocity tracking model reflects the error dynamics of the vehicle's longitudinal velocity.

4. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The adaptive adjustment of the speed tracking weights in the system output weighting matrix based on the vehicle's operating status specifically includes: An adaptive adjustment mechanism for speed tracking weights is designed based on the hyperbolic tangent function. The speed tracking weights are dynamically adjusted according to the thresholds set for the vehicle's lateral speed, yaw rate, lateral error, and heading error, as well as the adjustment parameters of the adaptive function.

5. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The adaptive adjustment of the longitudinal force distribution weights in the control input weighting matrix based on wheel slip ratio, wheel vertical load, and road adhesion coefficient specifically includes: The slip ratio of the i-th tire is defined as a function of the ratio of the longitudinal velocity of the vehicle body to the product of the corresponding wheel speed and the tire radius; Based on the slip ratio, the longitudinal control force weight of the wheel is defined as an exponential function related to the slip ratio, wherein the weight maintains a basic value when the slip ratio is small, and increases exponentially with the slip ratio when it exceeds a preset threshold. The longitudinal force distribution weight is calculated based on the road adhesion coefficient and vertical load of the i-th wheel to obtain an initial weight, which takes into account both the vertical load and the road adhesion conditions.

6. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The application of control input constraints specifically includes: A rate limit is introduced for the control input at the first control moment to suppress abrupt changes in the control quantity; and upper and lower limits are set for the control input at all control moments.

7. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The applied system output constraints specifically include: The system output is given upper and lower limits, and a relaxation variable is introduced to soften the constraint boundary and improve the feasibility of the optimization problem. The relaxation variable is a one-dimensional unit column vector.

8. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The step of transforming the multi-objective optimization function and the constraints into a quadratic programming problem with relaxation terms for solution specifically includes: Construct an optimization variable combination vector that includes control inputs and slack variables; The objective function with relaxation terms is transformed into a standard quadratic programming form, which minimizes the objective function and includes a weight matrix H of the quadratic terms and a coefficient vector f of the linear terms. Apply upper and lower bound constraints to the optimization variables; and Linear and equality constraints are applied to the optimization variables, the linear constraints being determined by the constraint matrix and the constraint boundary vector.

9. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, In the data-driven lateral dynamics model, a generalized yaw moment is introduced. This generalized yaw moment is generated by the difference in driving torque between the left and right wheels, and its calculation formula is related to the longitudinal force components of the left front, right front, left rear, and right rear tires.

10. The method for driving path tracking and stability control of a distributed drive electric vehicle according to claim 1, characterized in that, The obtained longitudinal traction force of each drive wheel is converted into the output torque of the corresponding drive motor through a preset transformation relationship, thereby achieving a decoupled transition between the optimized quantity and the actual execution quantity.

Images