A human-machine control right game lane keeping robust control method and system

Abstract

The application discloses a human-machine control right game lane keeping robust control method and system, first introduces a driver steering model simulating the steering behavior of a driver, and adopts an immune genetic algorithm (IGA) to identify the parameters of the driver steering model, and establishes a driver-in-the-loop vehicle-road model; secondly, considering the uncertainty of the model structure and parameters, the strong nonlinearity of vehicle dynamics and other disturbance factors, an output feedback gamma optimal H ∞ robust controller is designed based on the T-S fuzzy control theory; and finally, considering the driver steering behavior, vehicle lateral deviation and other factors, a human-machine control right game strategy is designed to realize the smooth dynamic distribution of the driving control right.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to a robust lane-keeping control method and system for human-machine control power game. Background Technology

[0002] Issues such as driving decision-making in complex scenarios, new infrastructure construction, social ethics, and legislation related to the definition of rights and responsibilities have significantly hindered the commercialization of autonomous driving technology. In the current technological and application context, scholars both domestically and internationally have proposed the concept of co-driving intelligent vehicles, incorporating advanced human cognition of fuzzy and uncertain problems into the feedback loop to improve the overall intelligence level of human-machine co-driving systems. This is precisely a practical application of human-machine hybrid augmented intelligence theory in the field of intelligent driving. Currently, the unsafe transition and cooperative control between drivers and conditional / highly automated driving systems are hot and challenging research areas. Furthermore, how to combine driver skills with new intelligent vehicle technologies to design a more cooperative driver-in-the-loop co-driving system to achieve efficient human-machine collaboration still faces significant challenges.

[0003] Co-driving lane keeping systems are a typical application of human-machine co-driving, enabling the driver and intelligent assistance system to share vehicle control and collaboratively complete lane keeping tasks. Compared to traditional mechanical steering systems, X-By-Wire technology achieves complete decoupling between driver commands and control actions, allowing the intelligent assistance system to correct driver commands and providing technical support for human-machine collaborative control. [5] According to the human-machine collaborative control structure, it mainly includes series and parallel types.

[0004] The characteristic of the serial type is that the intelligent assistance system corrects the driver's input commands based on the vehicle's safety boundary conditions, and generally does not correct them when the driver's input commands do not violate the safety boundary conditions, thereby satisfying the driver's control intentions to the greatest extent.

[0005] The characteristic of parallel operation is that the driver and the intelligent assistance system each make their own control decisions, and then design a control allocation strategy to achieve human-machine collaborative control. Therefore, under this control structure, how to rationally and dynamically allocate control rights between the driver and the intelligent assistance system is crucial and directly affects the performance of collaborative driving.

[0006] Designing reasonable weight allocation strategies to reduce human-machine conflict remains a key research focus in this field. Summary of the Invention

[0007] This application provides a robust lane-keeping control method and system based on a human-machine control power game. Its advantage is that it effectively enhances the cooperation between the intelligent assistance system and the driver during co-driving by using a human-vehicle-road model with the driver in the loop.

[0008] The technical solution of this application is as follows:

[0009] On the one hand, this application provides a robust lane-keeping control method based on a human-machine control power game, comprising the following steps:

[0010] A vehicle-road model is established based on a linear two-degree-of-freedom vehicle extended model and an electric power steering system model.

[0011] Construct a driver steering model and identify the optimal parameters of the driver steering model through model parameter identification;

[0012] A closed-loop human-vehicle-road model is established based on the vehicle-road model and the driver steering model. Based on the TS fuzzy theory, a fuzzy model of the human-vehicle-road model is established.

[0013] Establish a game model of human-machine control to determine the control rights of the intelligent assistance system and the driver.

[0014] Furthermore, the vehicle-road model is as follows:

[0015]

[0016] In the formula, Let u be the state vector. v =T d +T a For model input, T d T a These represent the steering torques of the driver and the intelligent assistance system, respectively, and φ = ρ is the road curvature at the aiming point. For model output, y L L represents the lateral deviation between the vehicle's center of mass and the aiming point in the vehicle coordinate system. near For single-point aiming distance, v y Let v be the lateral velocity. x For longitudinal velocity, These are the yaw angles at the vehicle's center of gravity and the aiming point, respectively. For the yaw angle deviation, ω r Let δ be the yaw rate. f For the front wheel steering angle, C f C r , i, b, and y are the lateral stiffness of the front and rear wheels, respectively; a and b are the distances between the front and rear axles and the vehicle's center of gravity, respectively; i, b, and y are the lateral stiffness of the front and rear wheels, respectively. s For the steering gear ratio, η t J is the tire contact width with the ground. s B is the equivalent rotational inertia of the steering system. s I is the equivalent damping coefficient of the steering system. z Let Z be the moment of inertia of the vehicle about the z-axis, and the coefficient matrices satisfy:

[0017]

[0018]

[0019] G v =[0, 0, v] x ,0,0,0] T

[0020]

[0021] Furthermore, the driver steering model is as follows:

[0022]

[0023] In the formula, x d =[x1,x2,x3,x4,x5] T Let u be the state vector. d =[θ far ,θ near ,δ sw ] T Input to the model, For model output; where, For a distant viewpoint, L far Pre-aiming distance for distant viewpoint, L For a close-up viewpoint, θ des =0 represents the desired near angle of view, δ sw This refers to the actual steering wheel angle. The steering torque estimated for the driver steering model has the following coefficient matrices:

[0024]

[0025]

[0026] C d =[0,0,0,0,1]

[0027] Among them, K P τ represents the proportional gain of the visual prediction module. lead τ lag These represent the lead and lag time constants of the visual compensation module, respectively, K. C For the proportional gain of the visual compensation module, τ p K represents the reaction time of the delay element. D For the proportional gain of the sensing element, K G For the proportional gain of the action element, τ1, τ k1 τ k2 All are driver characteristic constants, τ N The time constant of the neuromuscular dynamics module;

[0028] The model parameter identification steps include:

[0029] The prediction error method is used to roughly identify the parameters of the driver steering model. The model parameters to be identified are regarded as an antibody. The reciprocal of the sum of squares of the deviations between the driver's actual input and the model's predicted output is selected as the fitness function. After iterative optimization by the IGA genetic algorithm, an optimal antibody is obtained.

[0030] Furthermore, by combining equations (1) and (2), a closed-loop human-vehicle-road model is established:

[0031]

[0032] In the formula, Let u = T be the state vector. a Input to the model, For the model output, each coefficient matrix satisfies

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] Based on TS fuzzy theory, a fuzzy model of the human-vehicle-road model shown in equation (3) is established:

[0039]

[0040] In the formula, A i B i G i C i D i E i , Let z represent the coefficient matrix of the i-th local linear subsystem, and z be the measurement. For fuzzy antecedents, membership function satisfy:

[0041]

[0042] Based on the concept of parallel distributed compensation, the output feedback fuzzy controller model is defined as follows:

[0043]

[0044] In the formula, ξ is the state vector, and u is the controller output. Dc This is the corresponding state coefficient matrix;

[0045] Define the augmented vector χ = [x T ,ξ T ] T By combining equations (4) and (5), a closed-loop control system can be established:

[0046]

[0047] In the formula, each coefficient matrix satisfies

[0048]

[0049]

[0050]

[0051]

[0052] The output of the fuzzy controller shown in style (5) is rewritten as follows:

[0053]

[0054] In the formula,

[0055] For the closed-loop control system described by equation (6), establish the infinite norm of the closed-loop transfer function from φ to y, which satisfies:

[0056]

[0057] In the formula, the scalar γ > 0 represents the disturbance suppression degree, and the value of γ will determine the ability of the control system to suppress disturbances;

[0058] According to Schur's complement lemma, the γ-suboptimal H described by equation (8) ∞ The output feedback control problem is transformed into a convex optimization problem with LMI constraints and a linear objective function:

[0059] minγ

[0060]

[0061]

[0062] In the formula, X, Y, It is a feasible solution to LMI, where I is the identity matrix and Ψ ij satisfy

[0063]

[0064] Where (*) represents the symmetric transpose, and

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] After obtaining a set of feasible solutions for the LMI above, the coefficient matrices of the fuzzy controller shown in equation (5) can be obtained:

[0074]

[0075]

[0076]

[0077]

[0078] In the formula, M and N are both non-singular matrices and satisfy MNT = I - XY. Once X and Y are determined, the full-rank matrices M and N can be obtained by singular value decomposition.

[0079] Furthermore, a game theory model for human-machine control:

[0080]

[0081] In the formula, w a w d These represent the control rights of the intelligent assistance system and the driver, respectively; β1, β2, and β3 are all design parameters. This represents the normalized steering torque. This represents the normalized vehicle lateral composite deviation, where, λ1 and λ2 are the normalized values ​​of the lateral deviation and yaw angle deviation at the vehicle's center of gravity, respectively. λ1 and λ2 are both design parameters and satisfy λ1 + λ2 = 1.

[0082] Furthermore, it also includes human-computer friendliness evaluation methods, including the construction of the following human-computer friendliness evaluation indicators:

[0083] (1) Intervention factors

[0084] Define intervention factors for intelligent driver assistance systems to quantify the degree of intervention by these systems during shared driving:

[0085]

[0086] In the formula, ΔT max ψ represents the maximum permissible torque difference between the intelligent assistance system and the driver, and ψ is the intervention factor of the intelligent assistance system; when ψ→0, it indicates that the human and machine goals tend to be consistent; when ψ→1, it indicates that the intervention of the intelligent assistance system becomes more severe.

[0087] (2) Consistency rate

[0088]

[0089] In the formula, η co t represents the consistency rate. co t represents the time when the steering torque Ta of the intelligent assistance system is in the same direction as the driver's steering torque Td. total Total driving time;

[0090] (3) Resistance rate

[0091]

[0092] In the formula, η rst Indicates the resistance rate, t rst This indicates the time when the direction of Ta is opposite to the direction of Td and Ta is less than Td;

[0093] (4) Conflict rate

[0094]

[0095] In the formula, η cft t represents the conflict rate. cft This indicates the time when Ta is greater than Td.

[0096] On the other hand, this application provides a robust lane-keeping control system for human-machine control power game, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is called by the processor, it executes the method described above.

[0097] In summary, the beneficial effects of this application are as follows:

[0098] 1. In order to better mimic the driver's steering behavior, an IGA-optimized driver steering model was designed. The driver-vehicle-road model in the loop can effectively enhance the cooperation between the intelligent assistance system and the driver during co-driving.

[0099] 2. Considering the uncertainties in model structure and parameters, and the strong nonlinearity of vehicle dynamics, a suboptimal H-type output feedback γ control system was designed based on TS fuzzy control theory. ∞ Robust controller with good disturbance suppression capability.

[0100] 3. Based on the driver's steering behavior and the vehicle's lateral deviation, a human-machine control game model was designed to realize the explicit expression of human-machine interaction effects and the effective management of human-machine control. Attached Figure Description

[0101] Figure 1 This is a schematic diagram of a vehicle-road reference model based on single-point preview;

[0102] Figure 2 This is a schematic diagram of a driver steering model;

[0103] Figure 3 This is a schematic diagram of a game model involving human-machine control. Detailed Implementation

[0104] The specific embodiments of this application are described in detail below with reference to the accompanying drawings.

[0105] Example: This application provides a robust lane-keeping control method based on a human-machine control power game, comprising the following steps:

[0106] A vehicle-road model is established based on a linear two-degree-of-freedom vehicle extended model and an electric power steering system model.

[0107] Construct a driver steering model and identify the optimal parameters of the driver steering model through model parameter identification;

[0108] A closed-loop human-vehicle-road model is established based on the vehicle-road model and the driver steering model. Based on the TS fuzzy theory, a fuzzy model of the human-vehicle-road model is established.

[0109] Establish a game model of human-machine control to determine the control rights of the intelligent assistance system and the driver.

[0110] The steps are based on the linear two-degree-of-freedom vehicle extended model and the electric power steering system model to establish the vehicle-road model. The linear two-degree-of-freedom vehicle extended model and the electric power steering system (EPS) model can be found in the reference WANG Xuan-yao, CHENG Yi. Lane departure avoidance by man-machine cooperative control based on EPS and ESP systems, Journal of Mechanical Science and Technology[J].33(6):2929-2940,2019.

[0111] The vehicle-road model is a vehicle-road reference model based on single-point preview, such as... Figure 1 As shown. Its form is:

[0112]

[0113] In the formula, Let u be the state vector. v =T d +T a For model input, T d T a These represent the steering torques of the driver and the intelligent assistance system, respectively, and φ = ρ is the road curvature (external disturbance) at the aiming point. For model output, y L L represents the lateral deviation between the vehicle's center of mass and the aiming point in the vehicle coordinate system. near For single-point aiming distance, v y Let v be the lateral velocity. x For longitudinal velocity (assuming the aiming range v) x (remain unchanged) These are the yaw angles at the vehicle's center of gravity and the aiming point, respectively. ω represents the yaw rate deviation (the rotation angle of the vehicle coordinate system). r Let δ be the yaw rate. f For the front wheel steering angle, C f C r , i, b, and y are the lateral stiffness of the front and rear wheels, respectively; a and b are the distances between the front and rear axles and the vehicle's center of gravity, respectively; i, b, and y are the lateral stiffness of the front and rear wheels, respectively. s For the steering gear ratio, η t J is the tire contact width with the ground. s B is the equivalent rotational inertia of the steering system. s I is the equivalent damping coefficient of the steering system. z Let be the moment of inertia of the vehicle about the z-axis, and let the coefficient matrices satisfy .

[0114]

[0115]

[0116] G v =[0, 0, v] x ,0,0,0] T

[0117]

[0118] To more realistically mimic driver steering behavior, this application constructs a driver steering model within the framework proposed by SENTOUH C, NGUYEN AT, BENLOUCIF MA, et al. Driver-automation cooperation oriented approach for shared control of lane keeping assist systems[J]. IEEE Transactions on ControlSystems Technology, 27(5):1962–1978, 2018. For example... Figure 2 As shown, the driver steering model mainly consists of three parts: a two-point pre-aiming system, a neuromuscular dynamics system, and model parameter identification.

[0119] Establish a driver steering model:

[0120]

[0121] In the formula, x d =[x1,x2,x3,x4,x5] T Let u be the state vector. d =[θ far ,θ near ,δ sw ] T Input to the model, This is the output of the model. Among them, For a distant viewpoint, L far Pre-aiming distance for distant viewpoint, For a close-up viewpoint, θ des =0 represents the desired near angle of view, δ sw This refers to the actual steering wheel angle. The steering torque estimated for the driver steering model, and the coefficient matrices satisfy the following conditions:

[0122]

[0123]

[0124] C d =[0,0,0,0,1]

[0125] Among them, K P τ represents the proportional gain of the visual prediction module. lead τ lag These represent the lead and lag time constants of the visual compensation module, respectively, K. C For the proportional gain of the visual compensation module, τ p K represents the reaction time of the delay element. D For the proportional gain of the sensing element, K G For the proportional gain of the action element, τ1, τ k1 τ k2 All are driver characteristic constants, τ N This is the time constant of the neuromuscular dynamics module.

[0126] Model parameter identification is the process of finding model parameters that meet the design objectives. The prediction error method (PEM) is typically used for a rough identification of driver steering model parameters. To avoid the optimization algorithm getting trapped in local minima, this paper employs the IGA (immune genetic algorithm) to enhance the global search capability of the optimization algorithm.

[0127] The model parameters to be identified are regarded as an antibody. The reciprocal of the sum of squares of the deviations between the driver's actual input and the model's predicted output is selected as the fitness function. After IGA iterative optimization, an optimal antibody is obtained, as shown in Table 1.

[0128] Table 1 Driver Steering Model Parameters

[0129]

[0130] It is not difficult to see that the parameter K C With τ lead The relatively large value indicates that the driver steering model pays sufficient attention to changes in the near-field perspective to ensure trajectory tracking accuracy.

[0131] By combining equations (1) and (2), a closed-loop driver-vehicle-road (DVR) model is established:

[0132]

[0133] In the formula, Let u = T be the state vector. a Input to the model, For the model output, each coefficient matrix satisfies

[0134]

[0135]

[0136]

[0137]

[0138]

[0139] Based on TS fuzzy theory, a fuzzy model of the co-driving type LKAS shown in equation (3) is established:

[0140]

[0141] In the formula, A i B i G i C i D i E i , Let z represent the coefficient matrix of the i-th local linear subsystem, and z be the measurement. For fuzzy antecedents, membership function satisfy

[0142]

[0143] Based on the concept of parallel distributed compensation (PDC) (see reference DONG J, YANG G HH) ∞ Controller Synthesis via Switched PDC Scheme for Discrete-Time TS Fuzzy Systems[J].IEEE Transactions on Fuzzy Systems,17(3):544-555,2009.), the output feedback fuzzy controller model is defined as follows:

[0144]

[0145] In the formula, ξ is the state vector, and u is the controller output. D c This is the corresponding state coefficient matrix.

[0146] Define the augmented vector χ = [x T ,ξ T ] TBy combining equations (4) and (5), a closed-loop control system can be established:

[0147]

[0148] In the formula, each coefficient matrix satisfies

[0149]

[0150]

[0151]

[0152]

[0153] The output of the fuzzy controller shown in style (5) is rewritten as follows:

[0154]

[0155] In the formula,

[0156] For the closed-loop control system described by equation (6), establish the infinite norm of the closed-loop transfer function from φ to y, which satisfies:

[0157]

[0158] In the formula, the scalar γ > 0 represents the disturbance suppression degree, and the value of γ will determine the ability of the control system to suppress disturbances.

[0159] According to Schur's complement lemma, the γ-suboptimal H described by equation (8) ∞ The output feedback control problem is transformed into a convex optimization problem with LMI constraints and a linear objective function:

[0160] minγ

[0161]

[0162]

[0163] In the formula, X, Y, It is a feasible solution to LMI, where I is the identity matrix and Ψ ij satisfy

[0164]

[0165] Where (*) represents the symmetric transpose, and

[0166]

[0167]

[0168]

[0169]

[0170]

[0171]

[0172]

[0173]

[0174] After obtaining a set of feasible solutions for the LMI above, the coefficient matrices of the fuzzy controller shown in equation (5) can be obtained:

[0175]

[0176]

[0177]

[0178]

[0179] In the formula, M and N are both non-singular matrices and satisfy MN T =I-XY, once X and Y are determined, the full-rank matrices M and N can be obtained through singular value decomposition.

[0180] Based on the driver's steering behavior and the vehicle's lateral deviation, this application designs a human-machine control power game model to assign weights to the steering torque input by the driver and the intelligent assistance system, so as to achieve an explicit expression of the human-machine interaction effect.

[0181]

[0182] In the formula, w a w d These represent the control rights of the intelligent assistance system and the driver, respectively; β1, β2, and β3 are all design parameters. This represents the normalized steering torque. This represents the normalized overall lateral deviation of the vehicle. Among them, λ1 and λ2 are the normalized values ​​of the lateral deviation and yaw angle deviation at the vehicle's center of gravity, respectively. λ1 and λ2 are both design parameters and satisfy λ1 + λ2 = 1.

[0183] like Figure 3 As shown, where β1 = 4, β2 = 3, and β3 = 0.45, with the overall deviation... As the steering wheel gets larger, the intelligent assistance system will gradually gain more control, and as the driver gradually becomes more involved in the steering task... At that time, w a The rate of increase is relatively lower. Specifically, when β3 = 0, it indicates that the game model does not consider the driver's steering behavior, but only adjusts the human-machine control based on the comprehensive deviation at the vehicle's center of gravity position.

[0184] To better quantify the degree of cooperation between the driver and the intelligent assistance system during shared driving, a human-machine friendliness evaluation method is also included, which comprises the following human-machine friendliness evaluation indicators:

[0185] (1) Intervention factors

[0186] Define intervention factors for intelligent assistance systems to quantify the degree of intervention by intelligent assistance systems in driving during co-driving.

[0187]

[0188] In the formula, ΔT max Let ψ represent the maximum permissible torque difference between the intelligent assistance system and the driver, and let ψ be the intervention factor of the intelligent assistance system. When ψ→0, it indicates that the goals of the human and machine tend to be consistent; when ψ→1, it indicates that the intervention of the intelligent assistance system becomes more severe.

[0189] (2) Consistency rate

[0190]

[0191] In the formula, η co t represents the consistency rate. co This indicates the steering torque Ta of the intelligent assistance system and the steering torque T of the driver. d When the directions are the same, t total Total driving time.

[0192] (3) Resistance rate

[0193]

[0194] In the formula, η rst Indicates the resistance rate, t rst T represents a Direction and T d opposite directions and T a Less than T d The time.

[0195] (4) Conflict rate

[0196]

[0197] In the formula, η cft t represents the conflict rate. cft T representsa "Defeat" T d The time.

[0198] This application also provides a robust lane-keeping control system for human-machine control power game, including a processor and a memory. The memory stores a computer program, and when the computer program is called by the processor, it executes the method described above.

[0199] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the inventive concept of this application, and these all fall within the protection scope of this application.

Claims

1. A robust lane-keeping control method based on a human-machine control power game, characterized in that, Including the following steps: A vehicle-road model is established based on a linear two-degree-of-freedom vehicle extended model and an electric power steering system model. Construct a driver steering model and identify the optimal parameters of the driver steering model through model parameter identification; A closed-loop human-vehicle-road model is established based on the vehicle-road model and the driver steering model. Based on the TS fuzzy theory, a fuzzy model of the human-vehicle-road model is established. Establish a human-machine control power game model to determine the control rights of the intelligent assistance system and the driver; Human-machine control power game model: (11) In the formula, , These respectively represent the intelligent assistance system and the driver's control. , , These are all design parameters. This represents the normalized steering torque. This represents the normalized vehicle lateral composite deviation, where, , These are the normalized values ​​of the lateral deviation and yaw angle deviation at the vehicle's center of gravity, respectively. , All are design parameters, and they meet the requirements. .

2. The robust lane-keeping control method based on human-machine control power game as described in claim 1, characterized in that, The vehicle-road model is as follows: (1) In the formula, For state vectors, Input to the model, , These are the steering torques for the driver and the intelligent assistance system, respectively. The road curvature at the target point. For model output, This represents the lateral deviation between the vehicle's center of mass and the aiming point in the vehicle coordinate system. This is the single-point aiming distance. For lateral velocity, For longitudinal velocity, , These are the yaw angles at the vehicle's center of gravity and the aiming point, respectively. For yaw angle deviation, The yaw rate is angular velocity. For the front wheel steering angle, , These are the lateral stiffness of the front and rear wheels, respectively. These are the distances between the front and rear axles and the vehicle's center of gravity, respectively. This is the steering gear ratio. The width of the tire in contact with the ground. The equivalent rotational inertia of the steering system, This is the equivalent damping coefficient of the steering system. Let Z be the moment of inertia of the vehicle about the z-axis, and the coefficient matrices satisfy: 。 3. The robust lane-keeping control method based on human-machine control power game as described in claim 2, characterized in that, The driver's steering model is as follows: (2) In the formula, For state vectors, Input to the model, For model output; where, For a distant viewpoint, Pre-aiming distance for distant viewpoint, For close-up viewing, To achieve a close-up perspective, This refers to the actual steering wheel angle. The steering torque estimated for the driver steering model has the following coefficient matrices: in, For the proportional gain of the visual prediction module, , These represent the lead and lag time constants of the visual compensation module, respectively. For the proportional gain of the visual compensation module, The reaction time of the delay stage, For the proportional gain of the sensing element, For the proportional gain of the motion segment, , , All are driver characteristic constants. The time constant of the neuromuscular dynamics module; The model parameter identification steps include: The prediction error method is used to roughly identify the parameters of the driver steering model. The model parameters to be identified are regarded as an antibody. The reciprocal of the sum of squares of the deviations between the driver's actual input and the model's predicted output is selected as the fitness function. After iterative optimization by the IGA genetic algorithm, an optimal antibody is obtained.

4. The robust lane-keeping control method based on human-machine control power game as described in claim 3, characterized in that, By combining equations (1) and (2), a closed-loop human-vehicle-road model is established: (3) In the formula, For state vectors, Input to the model, For the model output, each coefficient matrix satisfies ； Based on TS fuzzy theory, a fuzzy model of the human-vehicle-road model shown in equation (3) is established: (4) In the formula, , , , , , , , The coefficient matrix represents the i-th local linear subsystem. For measurement, For fuzzy antecedents, membership function satisfy: Based on the concept of parallel distributed compensation, the output feedback fuzzy controller model is defined as follows: (5) In the formula, For state vectors, For controller output, , , , This is the corresponding state coefficient matrix; Define augmented vectors By combining equations (4) and (5), a closed-loop control system is established: (6) In the formula, each coefficient matrix satisfies The output of the fuzzy controller shown in style (5) is rewritten as follows: (7) In the formula, , ; For the closed-loop control system described by equation (6), establish a system from... arrive The closed-loop transfer function has an infinite norm and satisfies: (8) In the formula, scalar Indicates the degree of disturbance suppression. The value will determine the control system's ability to suppress disturbances; According to Schur's complement lemma, the equation (8) described Second best H ∞ The output feedback control problem is transformed into a convex optimization problem with LMI constraints and a linear objective function: (9) In the formula, , , , , , It is a feasible solution for LMI. It is the identity matrix. satisfy Where (*) represents the symmetric transpose, and After obtaining a set of feasible solutions for the LMI above, the coefficient matrices of the fuzzy controller shown in equation (5) can be obtained: (10) In the formula, M and N are both non-singular matrices and satisfy MNT=I-XY. Once X and Y are determined, the full-rank matrices M and N can be obtained by singular value decomposition.

5. The robust lane-keeping control method based on human-machine control power game as described in claim 1, characterized in that, It also includes human-computer friendliness evaluation methods, including the construction of the following human-computer friendliness evaluation indicators: (1) Intervention factors Define intervention factors for intelligent driver assistance systems to quantify the degree of intervention by these systems during shared driving: (12) In the formula, This represents the maximum permissible torque difference between the intelligent driver assistance system and the driver. As an intervention factor for the intelligent auxiliary system; when When this occurs, it indicates that the goals of humans and machines are converging; when... This indicates that the intervention of the intelligent assistance system is becoming increasingly serious; (2) Consistency rate (13) In the formula, Indicates the consistency rate. This indicates the time when the steering torque Ta of the intelligent assistance system is in the same direction as the driver's steering torque Td. Total driving time; (3) Resistance rate (14) In the formula, Indicates the resistance rate. This indicates the time when the direction of Ta is opposite to the direction of Td and Ta is less than Td; (4) Conflict rate (15) In the formula, Indicates the conflict rate. This indicates the time when Ta is greater than Td.

6. A robust lane-keeping control system based on human-machine control power game, characterized in that, It includes a processor and a memory, the memory storing a computer program, which, when invoked by the processor, executes the method as described in any one of claims 1-5.

Images