A fault-tolerant steering control method for autonomous driving vehicles with dynamic safety constraints

Abstract

The present invention relates to a steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints, comprising: step S1: obtaining obstacle information and real-time vehicle status; step S2: calculating the distance between the vehicle and each obstacle, taking the obstacle closest to the vehicle as the first obstacle, and generating a trigger function value based on the distance between the first obstacle and the vehicle and a trigger function, wherein the trigger function is a continuous function with a fast switching value of 0 to 1; step S3: based on the trigger function value, the original vehicle error obtained is corrected in combination with an obstacle function to obtain a corrected vehicle error, wherein the vehicle error includes a vehicle lateral error and a vehicle heading angle error; step S4: controlling the vehicle steering based on the corrected vehicle error. Compared with the prior art, the present invention constructs a distance-based trigger function combined with an obstacle function, and realizes continuous and stable switching with or without safety constraints. It has the advantage of reducing the load of the control system while ensuring vehicle safety.

Description

Technical Field

[0001] The present invention relates to the field of vehicle autonomous driving, and in particular to a steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints. Background Art

[0002] Autonomous driving vehicle systems have the characteristics of strong coupling, time-varying parameters, and high nonlinearity, and the safety of autonomous driving motion control is difficult to guarantee. In addition, system uncertainty, hardware failures, and complex external environments pose potential threats to their safety. Therefore, achieving high-reliability safety control of autonomous driving vehicles is a task of great practical significance and challenge.

[0003] In the process of motion control of autonomous vehicles, the steering system must provide lateral force to provide a suitable heading angle for avoiding obstacles and achieving path tracking. However, when the steering system fails, there is a deviation between the input steering angle of the control system and the actual steering angle, which will lead to a decrease in control performance. In severe cases, the vehicle may collide or cross the line, causing safety accidents and property losses. In the real environment, the uncertainty of the system and external interference will inevitably affect the accuracy of vehicle control. Therefore, the uncertainty of the system, external disturbances and steering actuator failures must be considered in the controller design to improve the tracking control accuracy of the vehicle, which can improve the safety of the vehicle. However, when the tracking accuracy reaches a certain level, further improving the accuracy will only increase the burden and energy consumption of the control system, and has no practical significance. In fact, it is enough to keep the vehicle within the safe range, and a small tracking error will not endanger the safety of the vehicle. Considering that imposing safety constraints will increase the control burden of the system, imposing safety constraints on a global scale limits the flexibility of safety constraints. Therefore, safety constraints only need to be imposed in special scenarios to ensure vehicle safety, such as encountering obstacles or narrowing the drivable area, etc., where vehicle safety needs to be ensured. Safety constraints are not required in other situations.

[0004] In summary, considering the safety requirements for the implementation of autonomous driving technology, system uncertainty, external disturbances and steering system failures must be considered when designing the controller; in addition, how to design dynamic safety constraints to ensure that the vehicle travels in a safe area while minimizing the system load is also of great significance to improving vehicle safety. Summary of the invention

[0005] The purpose of the present invention is to provide a steering fault-tolerant control method, device and storage medium for an autonomous driving vehicle with dynamic safety constraints.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints, comprising:

[0008] Step S1: Obtain obstacle information and real-time vehicle status, wherein the vehicle status includes vehicle position and vehicle heading angle;

[0009] Step S2: Calculate the distance between the vehicle and each obstacle, take the obstacle closest to the vehicle as the first obstacle, and generate a trigger function value based on the distance between the first obstacle and the vehicle and the trigger function, wherein the trigger function is a continuous function with a fast switching value between 0 and 1. When the distance between the first obstacle and the vehicle is less than the safety distance, the value of the trigger function is 1; when the distance between the first obstacle and the vehicle is greater than the safety distance, the value of the trigger function is quickly switched to 0;

[0010] Step S3: Based on the trigger function value, the original vehicle error is corrected in combination with the obstacle function to obtain a corrected vehicle error, so as to ensure that the vehicle is within the safety constraint, wherein the vehicle error includes the vehicle lateral error and the vehicle heading angle error;

[0011] Step S4: performing vehicle steering fault-tolerant control based on the corrected vehicle tracking error in combination with the radial basis function neural network.

[0012] The trigger function is:

[0013]

[0014] Where: l (d) is the trigger function, d is the distance between the first obstacle and the vehicle, d sv is the safety distance, e is the natural base, m is the system order, and μ is the trigger parameter.

[0015] The process of obtaining the original vehicle error includes:

[0016] Acquire a reference trajectory, and extract a reference position in the reference trajectory based on the reference trajectory and the vehicle position;

[0017] In the reference trajectory, obtaining a reference heading angle of the reference position;

[0018] An original vehicle position error is obtained based on the reference position and the vehicle position, and an original vehicle heading angle error is obtained based on the reference heading angle and the vehicle heading angle.

[0019] The corrected vehicle error is:

[0020]

[0021] θ l (e 1l )=e 1l Λ l (d)

[0022] Where:l is the corrected vehicle error, Q fl is the upper limit of the safety constraint, Q gl is the lower limit of the safety constraint, e 1l is the original vehicle error, θ l (e 1l ) is the trigger auxiliary variable, Λ l (d) is the trigger function. When l is 1, it indicates the vehicle position. When l is 2, it indicates the vehicle heading angle.

[0023] The step S4 comprises:

[0024] Step S4-1: generating a virtual control variable based on the corrected vehicle error;

[0025] Step S4-2: obtaining a derivative of the corrected vehicle error based on the virtual control variable;

[0026] Step S4-3: Generate a control signal based on the corrected vehicle error and its derivative.

[0027] The mathematical expression of the virtual control variable is:

[0028]

[0029] z 1 =[ξ 1 ,ξ 2 ] T

[0030] Among them: α is a dummy control variable, is the coefficient matrix after the barrier function is derived, k 1 is the first positive gain parameter, z 1 is the corrected vehicle error matrix, is the derivative of the system reference value.

[0031] The derivative of the corrected vehicle error is:

[0032] z 2 =x 2 -α

[0033] Where: z 2 is the derivative of the corrected vehicle error, x 2 is the derivative of the original vehicle error.

[0034] The control signal is:

[0035]

[0036] Where: u is the control signal, is the pseudo-inverse of the second kinetic model parameters, A(x) is the first kinetic model parameters, is the optimal update value of the weight vector, Ξ(x) is the basis function vector, is the observed value of the disturbance, is the derivative of the filtered value of the virtual control variable, k 2 is the second positive gain parameter,

[0037] A steering fault-tolerant control device for an autonomous driving vehicle with dynamic safety constraints comprises a memory, a processor, and a program stored in the memory, wherein the processor implements the above-mentioned method when executing the program.

[0038] A storage medium stores a program, which implements the above method when executed.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. Use obstacle functions to describe safety constraints, design distance-based trigger functions to implement dynamic safety constraints, and ensure the continuity and stability during the switching of safety constraints, which can reduce the load of the control system and improve driving smoothness.

[0041] 2. The selected trigger function can decay quickly when the distance between the first obstacle and the vehicle is less than the safety distance, and can be derivable everywhere, thereby ensuring the stability of the system during the switching of constraints.

[0042] 3. By setting the obstacle function, the system error constraint is transformed into the system output constraint, and the vehicle is controlled within the safety constraint by correcting the system error to ensure vehicle safety.

[0043] 4. Use radial basis function neural network to compensate for system steering failure and uncertainty, construct nonlinear disturbance observer to compensate for neural network fitting error and external disturbance, and use backstepping method to construct steering fault-tolerant control law to improve the safety of autonomous driving vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic flow chart of the main steps of the method of the present invention;

[0045] Figure 2 It is a schematic diagram of the technical route of the present invention;

[0046] Figure 3 It is a schematic diagram of the vehicle tracking dynamics model;

[0047] Figure 4 The diagram is a diagram of the change curve of the trigger function when the safety distance is 10;

[0048] Figure 5 Schematic diagram of the simulation results of the steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints. DETAILED DESCRIPTION

[0049] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0050] A fault-tolerant steering control method for an autonomous vehicle with dynamic safety constraints, such as Figure 1 As shown, including:

[0051] Step S1: Obtain obstacle information and real-time vehicle status, wherein the vehicle status includes vehicle position and vehicle heading angle;

[0052] Step S2: Calculate the distance between the vehicle and each obstacle, take the obstacle closest to the vehicle as the first obstacle, and generate a trigger function value based on the distance between the first obstacle and the vehicle and the trigger function, wherein the trigger function is a continuous function with a fast switching value between 0 and 1. When the distance between the first obstacle and the vehicle is less than the safety distance, the value of the trigger function is 1; when the distance between the first obstacle and the vehicle is greater than the safety distance, the value of the trigger function is quickly switched to 0;

[0053] In this embodiment, the designed trigger function is:

[0054]

[0055] Where: l (d) is the trigger function, d is the distance between the first obstacle and the vehicle, d sv is the safety distance, e is the natural base, m is the system order, and μ is the trigger parameter.

[0056] In some scenarios of autonomous driving, safety constraints need to be set to ensure vehicle safety. There are two types of safety constraints: unconstrained and constrained. The key is to ensure that the constraints are only valid in the specified phase and to ensure that the controller smoothly transitions between the unconstrained and constrained phases. To solve this problem, the above trigger function is proposed. When there is no obstacle, d→∞, from Figure 4 It can also be seen that the trigger function is a continuous and differentiable function. This continuity can ensure the stability of the system during the switching of constraints.

[0057] Figure 4 The changes at different values ​​are shown. For , μ determines the convergence speed to 0. Larger values ​​lead to slower convergence to 0. Therefore, smaller values ​​are conducive to fast transition between non-zero and zero.

[0058] Step S3: Based on the trigger function value, the original vehicle error is corrected in combination with the obstacle function to obtain a corrected vehicle error, so as to ensure that the vehicle is within the safety constraint, wherein the vehicle error includes the vehicle lateral error and the vehicle heading angle error;

[0059] In this embodiment, the process of obtaining the original vehicle error includes:

[0060] Acquire a reference trajectory, and extract a reference position in the reference trajectory based on the reference trajectory and the vehicle position;

[0061] In the reference trajectory, obtain the reference heading angle of the reference position;

[0062] An original vehicle position error is obtained based on the reference position and the vehicle position, and an original vehicle heading angle error is obtained based on the reference heading angle and the vehicle heading angle.

[0063] The corrected vehicle error is:

[0064]

[0065] θ l (e 1l )=e 1l Λ l (d)

[0066] Where: l is the corrected vehicle error, Q fl is the upper limit of the safety constraint, Q gl is the lower limit of the safety constraint, e 1l is the original vehicle error, θ l (e 1l ) is the trigger auxiliary variable, Λ l (d) is the trigger function. When l is 1, it indicates the vehicle position. When l is 2, it indicates the vehicle heading angle.

[0067] Specifically, by defining the vehicle tracking error e 1 =x 1 -x d , where x d =[e yd ,e ψd ],e yd =0,e ψd = 0, so e 1 =x 1 =[e y ,e ψ ]=[e 11 ,e 12 ] Define trigger auxiliary variables:

[0068] θ l (e 1l)=e 1l Λ l (d)

[0069] In order to describe the safety constraints, the following barrier function is proposed:

[0070]

[0071] Where Q fl and Q gl denote the upper and lower limits of the safety constraints, ξ=[ξ 1 ,ξ 2 ] T When d>d sv When l (d) = 0, so θ l (e 1l )=0, then ξ l =e 1l , then the system is in an unconstrained state, when d≤d sv When l (d) = 1, θ l (e 1l )=e 1l ,but:

[0072]

[0073] If e 1l Approach Q fl or Q gl ,ξ l →∞, in the feedback control system, use ξ l Replace the original system error e 1l ,Due to the existence of feedback control mechanism, the error will not exceed the safety boundary.

[0074] Step S4: performing vehicle steering fault-tolerant control based on the corrected vehicle tracking error combined with a radial basis function neural network, including:

[0075] Step S4-1: Generate a virtual control variable based on the corrected vehicle error. The mathematical expression of the virtual control variable is:

[0076]

[0077] z 1 =[ξ 1 ,ξ 2 ] T

[0078] Among them: α is a dummy control variable, is the coefficient matrix after the barrier function is derived, k 1 is the first positive gain parameter, z 1is the corrected vehicle error matrix, is the derivative of the system reference value.

[0079] Step S4-2: Obtain the derivative of the corrected vehicle error based on the virtual control variable. The derivative of the corrected vehicle error is:

[0080] z 2 =x 2 -α

[0081] Where: z 2 is the derivative of the corrected vehicle error, x 2 is the derivative of the original vehicle error.

[0082] Step S4-3: Generate a control signal based on the corrected vehicle error and its derivative.

[0083] Specifically, for the control part, first, a vehicle dynamics model including steering actuator failure, uncertainty and external disturbance needs to be established, such as Figure 3 As shown in the figure, considering the system actuator failure, uncertainty and external disturbance, the vehicle tracking dynamics model is established:

[0084]

[0085] u f =uP(x,u),P(x,u)=u-ρu-γ

[0086] In which, the state variable x is defined as [x 1 ,x 2 ] T , where x 1 =[e y ,e ψ ] T , e y is the lateral error of the vehicle, specifically the lateral distance error between the center of gravity of the vehicle and the reference trajectory, e ψ is the vehicle heading angle error, specifically the error with the reference heading angle, A(x) and B(x) are the first dynamic model parameter and the second dynamic model parameter, ΔA(x) and ΔB(x) represent the uncertainty within the system, u is the control signal, representing the system input, and u f represents the input of the steering system after the fault, γ represents additive fault, ρ represents multiplicative fault, d(t) is the external disturbance, and in this patent, it is assumed that the external disturbance is continuous and bounded.

[0087] Among them, the first kinetic model parameters and the second kinetic model parameters are:

[0088]

[0089] Where: C af and C ar represents the cornering stiffness of each front and rear tire, δ represents the front wheel steering angle, I z is the moment of inertia of the vehicle about the z-axis, m 1 is the vehicle mass, V x is the longitudinal velocity of the vehicle, l f and l r are the distances from the vehicle's center of gravity to the front and rear axles, respectively.

[0090] Integrating all the difficult variables in the dynamic model together yields:

[0091] G(x,u)=ΔA(x)-(B(x)+ΔB(x))P(x,u)+ΔB(x)u

[0092] Then radial basis function neural networks (RBFNNs) are used to compensate for actuator faults and uncertainties, and a nonlinear disturbance observer is used to observe and compensate for the neural network's fitting errors and external disturbances.

[0093] RBFNNs have universal approximation capabilities and show strong generalization and adaptability in dynamic uncertain environments. In this application, RBFNNs are used to approximate G(x,u) that includes system uncertainty and actuator failures:

[0094] G(x,u)=ω T Ξ(x)+τ(x),

[0095] in represents the RBFNN input vector, and τ(x) represents the NN approximation error. ω=[ω 1 ,ω 2 ,...,ω N ] T and Ξ(x)=[Ξ 1 (x),Ξ 2 (x),...,Ξ N (x)] T Represent the weight vector and basis function vector respectively, and N represents the number of nodes in the neural network. Gaussian function is selected as the radial basis function:

[0096]

[0097] where c i =[c 1i ,c 2i ,...c qi ] T and b iRepresent the width vector and the center vector respectively. In order to minimize the approximation error, it is necessary to find an optimal weight ω through iteration. Since RBFNN is a universal approximator, it can approximate any continuous function on a compact set with arbitrary accuracy, so, and is an unknown small positive number, and r is a positive real number.

[0098] In order to compensate for the neural network fitting error and external disturbance, the following disturbance composite function is defined:

[0099] D(t)=τ(x)+d(t)

[0100] Since the external disturbance is bounded, and but The kinetic equation can be rewritten as:

[0101]

[0102] Define the error variable of the system:

[0103] z 2 =x 2 -α,

[0104] Where α is a dummy control variable,

[0105] In order to estimate the composite disturbance function, a nonlinear disturbance observer is designed:

[0106]

[0107] is an auxiliary function and λ is a positive constant.

[0108] In this application, the barrier function is derived to obtain:

[0109]

[0110] Define z 1 =ξ, then:

[0111]

[0112] The dummy control variable α is designed as:

[0113]

[0114] where k 1 is the positive gain parameter of the design.

[0115] In order to reduce the impact of noise on the system, the virtual input parameter α is first-order filtered:

[0116]

[0117] Where χ represents the design constant of the filter and ρ is defined as the filtering error,

[0118]

[0119] The adaptive neural network fault-tolerant control law with dynamic safety constraints designed by backstepping method is as follows:

[0120]

[0121] Where: u is the control signal, is the pseudo-inverse of the second kinetic model parameters, A(x) is the first kinetic model parameters, is the optimal update value of the weight vector, Ξ(x) is the basis function vector, is the observed value of the disturbance, is the derivative of the filtered value of the virtual control variable, k 2 is the second positive gain parameter, The update rule is designed as

[0122]

[0123] where h is a positive constant, Λ W Indicates constant gain.

[0124] Finally, the boundedness of all control signals in the closed-loop system can be verified by using the Lyapunov method. The convergence and rationality of the weight update law and control input are proved. Figure 5 The simulation results also prove this point. Within the constraint activation range, the steering fault-tolerant control method (target control law) with dynamic safety constraints ensures that the vehicle travels within the safety boundary, effectively improving the safety of autonomous driving vehicles.

[0125] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

Claims

1. A steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints, characterized in that: include: Step S1: Obtain obstacle information and real-time vehicle status, wherein the vehicle status includes vehicle position and vehicle heading angle; Step S2: Calculate the distance between the vehicle and each obstacle, take the obstacle closest to the vehicle as the first obstacle, and generate a trigger function value based on the distance between the first obstacle and the vehicle and the trigger function; Step S3: Based on the trigger function value, the original vehicle error is corrected in combination with the obstacle function to obtain a corrected vehicle error, so as to ensure that the vehicle is within the safety constraint, wherein the vehicle error includes the vehicle lateral error and the vehicle heading angle error; Step S4: performing vehicle steering fault-tolerant control based on the corrected vehicle error combined with a radial basis function neural network; The trigger function is: in: To trigger the function, d is the distance between the first obstacle and the vehicle, d sv For a safe distance, e is the natural base, m is the system order, μ is the trigger parameter; The corrected vehicle error is: in: is the corrected vehicle error, is the upper limit of the safety constraint, is the lower limit of the safety constraint, is the original vehicle error, To trigger the auxiliary variable, To trigger the function, l When it is 1, it indicates the vehicle position; when it is 2, it indicates the vehicle heading angle.

2. The steering fault-tolerant control method of an autonomous driving vehicle with dynamic safety constraints according to claim 1, characterized in that: The process of obtaining the original vehicle error includes: Acquire a reference trajectory, and extract a reference position in the reference trajectory based on the reference trajectory and the vehicle position; In the reference trajectory, obtaining a reference heading angle of the reference position; An original vehicle position error is obtained based on the reference position and the vehicle position, and an original vehicle heading angle error is obtained based on the reference heading angle and the vehicle heading angle.

3. The steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints according to claim 1, characterized in that: The step S4 comprises: Step S4-1: generating a virtual control variable based on the corrected vehicle error; Step S4-2: obtaining a derivative of the corrected vehicle error based on the virtual control variable; Step S4-3: Generate a control signal based on the corrected vehicle error and its derivative.

4. The steering fault-tolerant control method for an autonomous driving vehicle with dynamic safety constraints according to claim 3, characterized in that: The mathematical expression of the virtual control variable is: in: is a dummy control variable, is the coefficient matrix after the barrier function is differentiated, k 1 is the first positive gain parameter, z 1 is the corrected vehicle error matrix, is the derivative of the system reference value, is the corrected vehicle position error, is the corrected vehicle heading angle error.

5. The steering fault-tolerant control method of an autonomous driving vehicle with dynamic safety constraints according to claim 4, characterized in that: The derivative of the corrected vehicle error is: in: is the derivative of the corrected vehicle error, is the derivative of the original vehicle error.

6. The steering fault-tolerant control method of an autonomous driving vehicle with dynamic safety constraints according to claim 5, characterized in that: The control signal is: in: u is the control signal, is the pseudo-inverse of the second kinetic model parameters, is the first kinetic model parameter, is the optimal update value of the weight vector, is the basis function vector, is the observed value of the disturbance, is the derivative of the filtered value of the virtual control variable, is the second positive gain parameter.

7. The steering fault-tolerant control method of an autonomous driving vehicle with dynamic safety constraints according to claim 1, characterized in that: The vehicle is a four-wheel vehicle.

Images