Path tracking optimization control method considering uncertainty of vehicle model

Abstract

The invention belongs to the technical field of intelligent vehicle control, and discloses a path tracking optimization control method considering the uncertainty of a vehicle model, which comprises the following steps of: establishing a two-degree-of-freedom dynamic model and analyzing the two-degree-of-freedom dynamic model to obtain geometric constraint conditions meeting vehicle movement and driving; constructing an objective function by using a multi-objective optimization strategy, and obtaining a state-space equation; constructing a prediction control model based on the state-space equation, performing optimal solution, generating an optimal control sequence, and inputting the optimal control sequence into the intelligent vehicle to control the vehicle to track a target trajectory; optimizing weighting matrix parameters of the transverse path tracking controller based on a genetic particle swarm optimization algorithm; and simulation is carried out and validity is verified. By adopting the path tracking optimization control method considering the uncertainty of the vehicle model, the use condition of the path tracking controller of the automatic driving vehicle is expanded, the driving safety of the vehicle in a complex traffic environment is improved, and the method has the characteristics of feasibility and stability.

Description

Technical Field

[0001] The present invention relates to the field of intelligent vehicle control technology, and in particular to a path tracking optimization control method considering vehicle model uncertainty. Background Art

[0002] During driving, the driver uses visual feedback to understand the road conditions and control the driving direction. In the actual driving process, the driver will be disturbed by many external factors and has instability. The shortcomings of this traditional vehicle driving method are becoming increasingly prominent, resulting in frequent traffic accidents. In response to the current traffic safety problems, the emergence of smart vehicles has given people hope to effectively solve these problems. Smart vehicles perceive the surrounding environment through the sensors equipped on the vehicle, avoiding the attenuation of the vehicle's path tracking ability under extreme conditions, which makes the car's ability to drive autonomously safer and more reliable.

[0003] Adhesion limit conditions are one of the main causes of fatal traffic accidents. Vehicles are prone to entering adhesion limit conditions when cornering at high speed, being hit sideways, driving on icy roads, or avoiding obstacles in an emergency. Under adhesion limit conditions, since the force required to maintain vehicle motion reaches or even exceeds the maximum adhesion that the road surface can provide, the vehicle is prone to instability and loss of control, causing serious traffic accidents. Existing vehicle path control methods for autonomous vehicles are usually designed based on a linear two-degree-of-freedom vehicle dynamics model. The designed path tracking controller is suitable for conditions where the vehicle's lateral acceleration is less than 0.4g and the tire's side slip characteristics are in the linear range. It does not cover conditions where the tire's side slip characteristics reach the saturation zone, which means that it cannot meet the needs of path tracking control under adhesion limit conditions.

[0004] Therefore, in view of the defects of the prior art, the present invention proposes a path tracking optimization control method taking into account the uncertainty of the vehicle model. Summary of the invention

[0005] The purpose of the present invention is to provide a path tracking optimization control method that takes into account the uncertainty of vehicle model. For front-wheel steering and rear-wheel drive vehicles, it meets the needs of vehicle path tracking control when the rear axle tire force reaches the adhesion limit, and is suitable for vehicle path tracking control under adhesion limit conditions, thereby expanding the use conditions of the path tracking controller of autonomous driving vehicles, improving the driving safety of vehicles in complex traffic environments, achieving strong constraint processing capabilities, and being feasible and stable.

[0006] To achieve the above object, the present invention provides a path tracking optimization control method considering vehicle model uncertainty, comprising the following steps:

[0007] S1. Establish a two-degree-of-freedom dynamics model based on the intelligent vehicle parameters and dynamic characteristics of the real vehicle and analyze it to obtain geometric constraints that meet the vehicle's motion and driving conditions;

[0008] S2, construct the objective function using a multi-objective optimization strategy, transform the nonlinear dynamics model into a linear model, and obtain the state space equation;

[0009] S3. Construct a predictive control model based on the state space equation, combine the objective function and constraints to find the optimal solution, generate the optimal control sequence, and input it into the intelligent vehicle to control the vehicle to track the target trajectory;

[0010] S4. Optimize the weighted matrix parameters of the lateral path tracking controller based on the genetic particle swarm optimization algorithm, design the fitness functions of the lateral error, heading error and front wheel angle to ensure the vehicle optimization effect;

[0011] S5. Perform simulation and verify the effectiveness of vehicle path tracking accuracy.

[0012] Preferably, in step S1, a two-degree-of-freedom dynamic model of lateral and yaw is established according to the physical intelligent vehicle structure, as shown below:

[0013]

[0014] Among them, v x 、v y are the longitudinal speed and the lateral speed respectively; is the lateral acceleration of the vehicle; ω is the yaw rate of the vehicle; is the vehicle yaw angular acceleration; I z is the moment of inertia of the vehicle around the Z axis; δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle; a and b are the distances from the front axle and rear axle to the center of mass respectively.

[0015] Preferably, in step S1, geometric constraints satisfying vehicle motion are constructed based on the derivation of geometric relationships, and the specific process is as follows:

[0016] S121. When the autonomous driving vehicle tracks the desired path, there are two errors: lateral error and heading error. The difference between the actual yaw angle of the vehicle and the desired heading angle, i.e., the vehicle heading deviation and lateral error rate, are calculated as follows:

[0017]

[0018] in, is the vehicle heading error change rate, is the lateral error change rate, is the heading angular velocity, is the reference heading angular velocity change rate; is the lateral velocity of the vehicle; v x is the vehicle longitudinal speed; is the vehicle heading angle error;

[0019] S122, establish an error dynamics model err with the vehicle lateral error, lateral error rate and heading error, heading error rate as state variables, and write the error dynamics model in the form of a state space equation, as shown below:

[0020]

[0021] in, e d is the lateral error; is the lateral error change rate; is the vehicle heading error, is the rate of change of vehicle heading error, u = δ f ; The matrices A, B, and C are as follows:

[0022]

[0023] Where m is the vehicle mass; ω is the vehicle yaw rate; I z is the moment of inertia of the vehicle around the Z axis; a and b are the distances from the front and rear axles to the center of mass, respectively; v x 、v y are the longitudinal velocity and the lateral velocity respectively; δ f is the front wheel turning angle; K f is the front axle cornering stiffness; K r is the lateral stiffness of the vehicle's rear axle.

[0024] Preferably, in step S2, based on the state quantity deviation and control quantity optimization of the vehicle, a multi-objective optimization strategy is used to construct an objective function, as shown below:

[0025]

[0026] Among them, ρ is the relaxation weight coefficient, ε is the relaxation factor to prevent misunderstanding, and N p 、N c are the prediction step length and control step length respectively; Q and R are the weight coefficient matrices of path tracking and control amount; η(t+i|t) represents the lateral position of the vehicle in the prediction time domain; η ref (t+i|t)) represents the lateral position of the vehicle corresponding to the reference path; Δu(t+i|t) represents the front wheel steering angle control increment.

[0027] Preferably, in step S2, a linear time-varying tire model is established, and the nonlinear tire model is linearized using a Taylor series expansion form. The equation of the tire cornering force after linearization is as follows:

[0028]

[0029] Among them, F yf and F yr are the front and rear wheel lateral forces calculated by the linear time-varying tire model; α f is the front wheel slip angle, α r is the rear wheel slip angle; and are the local slopes of the longitudinal forces of the front and rear wheels at the current slip angle, respectively, which are called the nominal cornering stiffness; F yf0 and F yr0 Defined as residual lateral force.

[0030] Preferably, in step S2, the front and rear wheel cornering stiffnesses of the vehicle are calculated based on the linear time-varying tire model, and the nonlinear dynamic model is converted into a linear model to obtain the state space equation as follows:

[0031]

[0032] Among them, u(t) = δ is the front wheel turning angle, is a state variable.

[0033] Preferably, the forward Euler method is used to discretize the state space equation, and the state expression at time k obtained after discretization is:

[0034]

[0035] Among them, A d (k) = I + TA (t), B d (k) = TB(t); T is the control step size; I is the unit matrix.

[0036] Preferably, in step S3, a predictive control model is constructed based on the state space equation, and the specific process is as follows:

[0037] S311, using the state space equation to approximately linearize the nonlinear dynamic model, to obtain the state quantity after applying the control quantity;

[0038] S312, perform Taylor expansion calculation and discretization processing on the state quantity, and further obtain the state constraint equation based on the vehicle front wheel angle control constraint, as shown below:

[0039]

[0040] Among them, m is the number of control quantities; n is the number of state quantities; η(k) is the system output; is the state transfer matrix; is the control input matrix; Δu(k) is the control input change; η(k) is the output vector; is the output matrix; ξ(k) is the state vector;

[0041] S313, clarifying the relationship between variables according to the state constraint equation, and constructing a predictive control model using the output at the future moment;

[0042] Among them, the sequence of control quantities and states in the prediction time domain is as follows:

[0043]

[0044] According to the state constraint equation, the relationship between variables is clarified, and the output at future moments is used to construct a predictive control model, as shown below:

[0045] Y(k)=ψ(k)ξ(k)+Θ(k)ΔU(k)

[0046] Among them, ψ(k) is the state quantity in the control time domain, which is the Jacobian matrix relative to ξ; Θ(k) is the control increment in the control time domain, which is the Jacobian matrix relative to U.

[0047] Preferably, in step S4, based on the particle swarm optimization algorithm, the particle swarm velocity update formula is as follows:

[0048]

[0049] Among them, X i (t) is the position of the particle at the tth second; V i (t) is the velocity of the particle in the tth second; P b is the individual optimal value particle; G b is the global optimal value particle; C1 and C2 are both learning factors;

[0050] The crossover operation in the genetic algorithm is integrated into the standard particle swarm algorithm. The particles will be hybridized in pairs. The speed and position update formula of the hybridized particles are as follows:

[0051]

[0052] Among them, P c is the genetic crossover probability; V1' and V2' are the updated particle velocities.

[0053] Preferably, in step S4, the fitness function J expression is designed as follows:

[0054]

[0055] Among them, ω1, ω2, ω3 are Δe d , and Δδ f The weight coefficient of the fitness function. The larger the weight coefficient, the higher the importance of control. d , Δδ f are the RMS values ​​of vehicle lateral error, heading error and front wheel steering angle respectively.

[0056] Therefore, the present invention adopts the above-mentioned path tracking optimization control method that takes into account the uncertainty of the vehicle model. It is aimed at front-wheel steering and rear-wheel drive vehicles, and meets the needs of vehicle path tracking control when the rear axle tire force reaches the adhesion limit. It is suitable for vehicle path tracking control under adhesion limit conditions, thereby expanding the use conditions of the path tracking controller of the autonomous driving vehicle, improving the driving safety of the vehicle in complex traffic environments, achieving strong constraint processing capabilities, and being feasible and stable.

[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flow chart of a path tracking optimization control method considering vehicle model uncertainty of the present invention;

[0059] Figure 2 is a vehicle dynamics error model diagram of the present invention;

[0060] Figure 3 It is the principle diagram of local linearization of tire force of the present invention;

[0061] Figure 4 is a diagram of a flow chart of the GA-PSO algorithm of the present invention;

[0062] Figure 5 It is the path tracking control optimization result diagram of the present invention; among them, (a) is the vehicle double lane change path diagram; (b) is the vehicle continuous lane change path diagram. DETAILED DESCRIPTION

[0063] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0064] like Figure 1 As shown, the present invention provides a path tracking optimization control method considering vehicle model uncertainty, comprising the following steps:

[0065] S1. Establish a two-degree-of-freedom dynamics model based on the intelligent vehicle parameters and dynamic characteristics of the real vehicle and analyze it to obtain geometric constraints that meet the vehicle's motion and driving conditions;

[0066] S2, construct the objective function using a multi-objective optimization strategy, transform the nonlinear dynamics model into a linear model, and obtain the state space equation;

[0067] S3. Construct a predictive control model based on the state space equation, combine the objective function and constraints to find the optimal solution, generate the optimal control sequence, and input it into the intelligent vehicle to control the vehicle to track the target trajectory;

[0068] S4. Optimize the weighted matrix parameters of the lateral path tracking controller based on the genetic particle swarm optimization algorithm, design the fitness functions of the lateral error, heading error and front wheel angle to ensure the vehicle optimization effect;

[0069] S5. Perform simulation and verify the effectiveness of vehicle path tracking accuracy.

[0070] Example

[0071] S1. Establish a two-degree-of-freedom dynamics model based on the intelligent vehicle parameters and dynamic characteristics of the real vehicle and analyze it to obtain geometric constraints that meet the vehicle's motion and driving conditions;

[0072] S11. According to the physical intelligent vehicle structure, an xoy coordinate system is established, and the x-axis is defined as the direction perpendicular to the axle and pointing to the vehicle's forward direction, and the y-axis is defined as the direction starting from the vehicle's center of mass and parallel to the axle; according to the front and rear kinematic constraints of the vehicle, a two-degree-of-freedom dynamic model of lateral and yaw is established, as shown below:

[0073]

[0074] Among them, v x 、v y are longitudinal speed and lateral speed (km / h), respectively; is the lateral acceleration of the vehicle; ω is the yaw rate of the vehicle (rad / s); is the vehicle yaw angular acceleration; I z is the moment of inertia of the vehicle around the Z axis (kg·m 2 );δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle (kg); a and b are the distances from the front axle and rear axle to the center of mass (m), respectively.

[0075] S12. Derive based on geometric relationships and construct geometric constraint conditions that satisfy the vehicle's motion.

[0076] S121. When an autonomous driving vehicle tracks a desired path, there are two types of errors: lateral error and heading error.

[0077] like Figure 2 As shown, the vehicle lateral distance deviation ed Refers to the shortest straight-line distance from the vehicle's center of mass to the desired path; heading deviation Refers to the deviation of the vehicle's heading angle from the desired road direction.

[0078] The difference between the actual yaw angle and the desired heading angle of the vehicle, i.e., the vehicle heading deviation and lateral error rate, are calculated as follows:

[0079]

[0080] in, is the vehicle heading error change rate, is the lateral error change rate, is the heading angular velocity, is the reference heading angular velocity change rate; is the lateral velocity of the vehicle; v x is the vehicle longitudinal speed; is the vehicle heading angle error;

[0081] S122. On conventional roads, the angular velocity of the vehicle heading changes slightly and can be ignored. The error dynamics model err is established with the vehicle lateral error, lateral error rate and heading error, heading error rate as state variables, and the error dynamics model is written in the form of state space equations as shown below:

[0082]

[0083] in, e d is the lateral error; is the lateral error change rate; is the vehicle heading error, is the rate of change of vehicle heading error, u = δ f ; The matrices A, B, and C are as follows:

[0084]

[0085] Where, m is the vehicle mass (kg); ω is the vehicle yaw rate (rad / s); I z is the moment of inertia of the vehicle around the Z axis (kg·m 2 ); a, b are the distances from the front axle and rear axle to the center of mass (m); v x 、v y are longitudinal speed and lateral speed (km / h), respectively; δ f is the front wheel turning angle (rad); K f is the front axle cornering stiffness; K r is the lateral stiffness of the vehicle's rear axle.

[0086] S2. Use multi-objective optimization strategy to construct the objective function, transform the nonlinear dynamic model into a linear model, and obtain the state space equation.

[0087] S21. Based on the state deviation and control quantity optimization of the vehicle, the objective function is constructed using a multi-objective optimization strategy as shown below:

[0088]

[0089] Among them, ρ is the relaxation weight coefficient, ε is the relaxation factor to prevent misunderstanding, and N p 、N c are the prediction step length and control step length respectively; Q and R are the weight coefficient matrices of path tracking and control amount; η(t+i|t) represents the lateral position of the vehicle in the prediction time domain; η ref (t+i|t)) represents the lateral position of the vehicle corresponding to the reference path; Δu(t+i|t) represents the front wheel steering angle control increment.

[0090] S22. When the wheel is in a nonlinear state, the fixed tire cornering stiffness value changes accordingly. A control system is designed to react to the tire force of the vehicle in a nonlinear state using a linear time-varying tire.

[0091] The nonlinear dynamic model analysis specifically includes: defining the vehicle's driving surface as flat and without vertical motion; defining the suspension device and the vehicle as rigid, and ignoring the effect of the coupling between the suspension and the tire on path tracking; considering the tire's side slip characteristics and ignoring the coupling relationship between the longitudinal and lateral directions of the tire; not considering the left-right transfer of loads and ignoring the load transfer between the front and rear axles when the vehicle is driving at low speeds; and ignoring the effects of the vehicle's lateral and longitudinal aerodynamics.

[0092] In addition, the constraints include kinematic constraints and dynamic constraints; specifically, they include center of mass sideslip angle constraints, vehicle adhesion condition constraints, and tire sideslip angle constraints.

[0093] Therefore, a linear time-varying tire model is established, and the nonlinear tire model is linearized using the Taylor series expansion form. The equation of the tire cornering force after linearization is as follows:

[0094]

[0095] Among them, F yf and F yr are the front and rear wheel lateral forces calculated by the linear time-varying tire model; α f is the front wheel slip angle, α r is the rear wheel slip angle; and are the local slopes of the longitudinal forces of the front and rear wheels at the current slip angle, respectively, which are called the nominal cornering stiffness; F yf0 and Fyr0 is defined as the residual lateral force. The principle of local linearization is as follows Figure 3 shown.

[0096] S23. Calculate the front and rear wheel cornering stiffness of the vehicle based on the linear time-varying tire model, convert the nonlinear dynamic model into a linear model, and obtain the state space equation as follows:

[0097]

[0098] Among them, u(t) = δ is the front wheel turning angle, is a state variable.

[0099] S24. The forward Euler method is used to discretize the above state space equation. After discretization, the state expression at time k is obtained as follows:

[0100]

[0101] Among them, A d (k) = I + TA (t), B d (k) = TB(t); T is the control step size; I is the unit matrix.

[0102] S3. Construct a predictive control model based on the state-space equation, combine the objective function and constraints to find the optimal solution, generate the optimal control sequence, and input it into the intelligent vehicle to control the vehicle to track the target trajectory.

[0103] S31. Construct a predictive control model based on the state space equation. The specific process is as follows:

[0104] S311, using the state space equation to approximately linearize the nonlinear dynamic model, to obtain the state quantity after applying the control quantity;

[0105] S312, perform Taylor expansion calculation and discretization processing on the state quantity, and further obtain the state constraint equation based on the vehicle front wheel angle control constraint, as shown below:

[0106]

[0107] Among them, m is the number of control quantities; n is the number of state quantities; η(k) is the system output; is the state transfer matrix; is the control input matrix; Δu(k) is the control input change; η(k) is the output vector; is the output matrix; ξ(k) is the state vector.

[0108] S313. Clarify the relationship between variables based on the state constraint equation and use the output at future times to build a predictive control model.

[0109] The MPC controller continuously infers the state of the system from the state of the next moment, so let the prediction interval be N p , the control interval is N c The sequence of control quantities and states in the prediction time domain is as follows:

[0110]

[0111] According to the state constraint equation, the relationship between variables is clarified, and the output at future moments is used to construct a predictive control model, as shown below:

[0112] Y(k)=ψ(k)ξ(k)+Θ(k)ΔU(k)

[0113] Among them, ψ(k) is the state quantity in the control time domain, which is the Jacobian matrix relative to ξ; Θ(k) is the control increment in the control time domain, which is the Jacobian matrix relative to U.

[0114] In the embodiment of the present invention, the initial prediction time domain N is defined p =30, N c = 20. Obtain vehicle and environment information, including the vehicle's current location information and speed information, reference path information, and expected speed information, and use them as input to establish a prediction model.

[0115] S32. The predictive control model compares the output measurement value with the model prediction value to obtain the prediction error of the predictive control model; the prediction error is used to correct and iteratively optimize the model prediction value until an error-free prediction value is output.

[0116] S33, the prediction model is optimized and solved according to the optimized prediction value in combination with the objective function and the constraint conditions to obtain the optimal control sequence in the control time domain; the optimal control sequence is input into the intelligent vehicle, the steering of the vehicle is controlled according to the current control amount, and the target trajectory is tracked.

[0117] S4. Based on the genetic particle swarm optimization algorithm, the weighted matrix parameters of the lateral path tracking controller are optimized, and the fitness functions of the lateral error, heading error and front wheel angle are designed to ensure the vehicle optimization effect.

[0118] S41, based on the particle swarm optimization algorithm (PSO), with a group of random particles as the initial point, with iterative evolution, the particle speed V that moves towards the optimal solution is continuously generated i (t+1). In this process, particles are subjected to local optimal and global optimal effects and search for solutions in the search space.

[0119] In each iteration, particles track individual extreme values ​​(P b ) and group extreme value (Gb ) to update its position. The moving speed of the particle is mainly determined by the current speed, individual extreme value and individual extreme value. Once the new speed is determined, the new position of the particle can be calculated. The particle group speed update formula is as follows:

[0120]

[0121] Among them, X i (t) is the position of the particle at the tth second; V i (t) is the velocity of the particle in the tth second; P b is the individual optimal value particle; G b is the global optimal value particle; C1 and C2 are both learning factors.

[0122] S42. In order to enhance the global search capability of the particle swarm optimization algorithm and prevent particles from falling into the local optimum, the crossover operation in the genetic algorithm is integrated into the standard particle swarm algorithm to increase the diversity of the particle population.

[0123] Drawing on the evolutionary idea of ​​genetic algorithms, a part of the particles in the population is randomly selected with a certain probability P c Perform crossover operation, the GA-PSO algorithm flow chart is as follows Figure 4 As shown. These selected particles will be hybridized in pairs, and the speed and position update formula of the hybridized particles are as follows:

[0124]

[0125] Among them, P c is the genetic crossover probability; V1' and V2' are the updated particle velocities.

[0126] S43. The objective function of the GA-PSO optimization algorithm design is to meet the conditions that the lateral error and heading error should be as small as possible, and the energy consumption requirement of the control is relatively small (the front wheel turning angle is small).

[0127] The system's lateral and longitudinal deviations and front wheel steering angles are of different orders of magnitude, so the three variables are processed in a unified dimension. The fitness function J contains three variables, and the fitness function J expression is as follows:

[0128]

[0129]

[0130] Among them, ω1, ω2, ω3 are Δe d , and Δδ f The weight coefficient of the fitness function. The larger the weight coefficient, the higher the importance of control. d , Δδ f are the RMS values ​​of vehicle lateral error, heading error and front wheel steering angle respectively.

[0131] S5. Perform simulation and verify the effectiveness of vehicle path tracking accuracy.

[0132] like Figure 5 As shown in the figure, it can be seen from the vehicle path tracking control optimization result diagram that the three controllers, MPC control, GAPSO-MPC control and GAPSO-LTV-MPC control, can all track the desired path, but the traditional fixed-value MPC path tracking controller has the worst effect and cannot guarantee the path tracking accuracy of the intelligent vehicle under different working conditions.

[0133] On a double lane change road, GAPSO-MPC control, due to the influence of the optimization algorithm, further improves the path tracking accuracy. The best intelligent vehicle tracking accuracy is GAPSO-LTV-MPC control, which establishes a linear time-varying vehicle model, calculates the vehicle's stiffness change and introduces it into the MPC path tracking control calculation. The final path tracking accuracy is the highest, and GAPSO-LTV-MPC control vehicles have good tracking effects on different roads including continuous shifting conditions.

[0134] Therefore, the present invention adopts the above-mentioned path tracking optimization control method considering the uncertainty of the vehicle model, and designs a control system for the tire force of the linear time-varying tire to react to the vehicle in the nonlinear state according to the changes in the curvature and heading angle of the reference path of the intelligent vehicle; establishes a linear time-varying tire model, linearizes the nonlinear tire model in the form of Taylor series expansion, and readjusts the expected front wheel turning angle of the path tracking motion of the intelligent vehicle according to the linear time-varying vehicle model. Based on the size of the lateral error and the heading error during path tracking, a genetic particle swarm optimization algorithm controller is designed to realize the search for the optimal value of the MPC tracking algorithm prediction time domain. Finally, the MPC path tracker output control quantity is designed to realize the tracking of the reference path, and the path tracking is performed by selecting a lower speed and a smaller prediction time domain when the curvature is large. In the case of a small curvature, a higher speed and a larger prediction time domain are selected for path tracking, thereby improving the algorithm adaptability of the MPC path tracking to different reference paths. The present invention expands the use conditions of the path tracking controller of the autonomous driving vehicle, improves the driving safety of the vehicle in a complex traffic environment, meets the control requirements and stability of existing intelligent vehicles, and achieves a strong constraint processing capability, with high feasibility and stability.

[0135] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A path tracking optimization control method considering vehicle model uncertainty, characterized in that: The following steps are involved: S1. Establish a two-degree-of-freedom dynamics model based on the intelligent vehicle parameters and dynamic characteristics of the real vehicle and analyze it to obtain geometric constraints that meet the vehicle's motion and driving conditions; S2, construct the objective function using a multi-objective optimization strategy, convert the nonlinear dynamics model into a linear model, and obtain the state space equation; S3. Construct a predictive control model based on the state space equation, combine the objective function and constraints to find the optimal solution, generate the optimal control sequence, and input it into the intelligent vehicle to control the vehicle to track the target trajectory; S4. Optimize the weighted matrix parameters of the lateral path tracking controller based on the genetic particle swarm optimization algorithm, design the fitness functions of the lateral error, heading error and front wheel angle to ensure the vehicle optimization effect; S5. Perform simulation and verify the effectiveness of vehicle path tracking accuracy.

2. A path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S1, a two-degree-of-freedom dynamic model of lateral and yaw is established according to the physical intelligent vehicle structure, as shown below: Among them, v x 、v y are the longitudinal speed and the lateral speed respectively; is the lateral acceleration of the vehicle; ω is the yaw angular velocity of the vehicle; is the vehicle yaw angular acceleration; I z is the moment of inertia of the vehicle around the Z axis; δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle; a and b are the distances from the front axle and rear axle to the center of mass respectively.

3. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S1, geometric constraints that satisfy the vehicle's motion are constructed based on the derivation of geometric relationships. The specific process is as follows: S121. When the autonomous driving vehicle tracks the desired path, there are two errors: lateral error and heading error. The difference between the actual yaw angle of the vehicle and the desired heading angle, i.e., the vehicle heading deviation and lateral error rate, are calculated as follows: in, is the vehicle heading error change rate, is the lateral error change rate, is the heading angular velocity, is the reference heading angular velocity change rate; is the lateral velocity of the vehicle; v x is the vehicle longitudinal velocity; is the vehicle heading angle error; S122, establish an error dynamics model err with the vehicle lateral error, lateral error rate and heading error, heading error rate as state variables, and write the error dynamics model in the form of a state space equation, as shown below: in, e d is the lateral error; is the lateral error change rate; is the vehicle heading error, is the rate of change of vehicle heading error, u = δ f ; The matrices A, B, and C are as follows: Where m is the vehicle mass; ω is the vehicle yaw rate; I z is the moment of inertia of the vehicle around the Z axis; a and b are the distances from the front and rear axles to the center of mass, respectively; v x 、v y are the longitudinal velocity and the lateral velocity respectively; δ f is the front wheel turning angle; K f is the front axle cornering stiffness; K r is the lateral stiffness of the vehicle's rear axle.

4. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S2, based on the state deviation and control quantity optimization of the vehicle, a multi-objective optimization strategy is used to construct an objective function as shown below: Among them, ρ is the relaxation weight coefficient, ε is the relaxation factor to prevent misunderstanding, and N p 、N c are the prediction step length and control step length respectively; Q and R are the weight coefficient matrices of path tracking and control amount; η(t+i|t) represents the lateral position of the vehicle in the prediction time domain; η ref (t+i|t)) represents the lateral position of the vehicle corresponding to the reference path; Δu(t+i|t) represents the front wheel steering angle control increment.

5. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S2, a linear time-varying tire model is established, and the nonlinear tire model is linearized using the Taylor series expansion form. The equation of the tire cornering force after linearization is as follows: Among them, F yf and F yr are the front and rear wheel lateral forces calculated by the linear time-varying tire model; α f is the front wheel slip angle, α r is the rear wheel slip angle; and are the local slopes of the longitudinal forces of the front and rear wheels at the current slip angle, respectively, which are called the nominal cornering stiffness; F yf0 and F yr0 Defined as residual lateral force.

6. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S2, the front and rear wheel cornering stiffness of the vehicle is calculated based on the linear time-varying tire model, and the nonlinear dynamic model is converted into a linear model to obtain the state space equation as follows: Among them, u(t) = δ is the front wheel turning angle, is a state variable.

7. The path tracking optimization control method considering vehicle model uncertainty according to claim 6 is characterized in that: The forward Euler method is used to discretize the state space equation. After discretization, the state expression at time k is obtained as follows: Among them, A d (k) = I + TA (t), B d (k) = TB(t); T is the control step size; I is the unit matrix.

8. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S3, a predictive control model is constructed based on the state space equation. The specific process is as follows: S311, using the state space equation to approximately linearize the nonlinear dynamic model, to obtain the state quantity after applying the control quantity; S312, perform Taylor expansion calculation and discretization processing on the state quantity, and further obtain the state constraint equation based on the vehicle front wheel angle control constraint, as shown below: Among them, m is the number of control quantities; n is the number of state quantities; η(k) is the system output; is the state transfer matrix; is the control input matrix; Δu(k) is the control input change; η(k) is the output vector; is the output matrix; ξ(k) is the state vector; S313, clarifying the relationship between variables according to the state constraint equation, and constructing a predictive control model using the output at the future moment; Among them, the sequence of control quantities and states in the prediction time domain is as follows: According to the state constraint equation, the relationship between variables is clarified, and the output at future moments is used to construct a predictive control model, as shown below: Y(k)=ψ(k)ξ(k)+Θ(k)ΔU(k) Among them, ψ(k) is the state quantity in the control time domain, which is the Jacobian matrix relative to ξ; Θ(k) is the control increment in the control time domain, which is the Jacobian matrix relative to U.

9. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S4, based on the particle swarm optimization algorithm, the particle swarm velocity update formula is as follows: Among them, X i (t) is the position of the particle at the tth second; V i (t) is the velocity of the particle in the tth second; P b is the individual optimal value particle; G b is the global optimal value particle; C1 and C2 are both learning factors; The crossover operation in the genetic algorithm is integrated into the standard particle swarm algorithm. The particles will be hybridized in pairs. The speed and position update formula of the hybridized particles are as follows: Among them, P c is the genetic crossover probability; V1' and V2' are the updated particle velocities.

10. The path tracking optimization control method considering vehicle model uncertainty according to claim 1, characterized in that: In step S4, the fitness function J expression is designed as follows: Among them, ω1, ω2, ω3 are Δe d , and Δδ f The weight coefficient of the fitness function. The larger the weight coefficient, the higher the importance of control. d , Δδ f are the RMS values ​​of vehicle lateral error, heading error and front wheel steering angle respectively.

Images