A safety control method, system and device for automatic steering of an unmanned vehicle

Abstract

The application relates to the technical field of automobile control, in particular to a safety control method, system and device for automatic steering of an unmanned vehicle, wherein, in the safety control method, in the case that an actuator attack occurs in the automatic steering control process of the unmanned vehicle, it is assumed that the actuator attack signal has no specific assumption rule, the accurate actual attack signal is obtained after nonlinear mapping and weighted summation processing in a neural network method, the actual control signal input under the actuator attack is constructed based on the actual attack signal, then the integral sliding mode strategy is used to design a sliding mode reaching law, the sliding mode controller signal is constructed in combination with the actual attack signal, and the sliding mode controller safety signal range is obtained in combination with the derivative of the designed Lyapunov function, so that the unmanned vehicle can be automatically and safely steered under the actuator attack, and the safety control of the automatic steering of the unmanned vehicle is realized.

Description

Technical Field

[0001] This invention relates to the field of automotive control technology, specifically to a safety control method, system, and device for automatic steering of unmanned vehicles. Background Technology

[0002] With the development of new technologies such as the Internet and artificial intelligence, automobiles have transformed from traditional, mechanical means of transportation into intelligent, connected new types of smart mobile terminals. Autonomous vehicles, equipped with advanced sensors, controllers, and actuators, transmit vehicle information in real time via a Controller Area Network (CAN bus). The emergence of autonomous vehicles has not only reduced the frequency of traffic accidents but also improved the road traffic environment, attracting widespread attention from numerous scholars.

[0003] While self-driving vehicles bring convenience, news about cybersecurity incidents involving them also frequently appears in the public eye. The main types of cyberattacks occurring during actual vehicle information transmission include spoofing attacks and denial-of-service attacks. Both types of attacks negatively impact safe vehicle operation. Only by ensuring the cybersecurity of self-driving vehicles, enabling them to maintain good control performance even under cyberattacks, can they truly gain public acceptance.

[0004] Traditional autonomous vehicle automatic steering control mostly achieves steering by controlling the front wheel angle. When an actuator attack occurs during automatic steering, a cyberattack can hijack the communication channel from the electronic control unit to the entire vehicle, modifying the control commands of the onboard controller. These modified commands are then sent to the vehicle's steering gear, resulting in poor steering performance and potentially causing accidents. Sliding mode variable structure control, due to its strong robustness, is widely used in autonomous vehicle automatic steering control. However, when an actuator attack occurs, the sliding mode controller requires a large switching gain to suppress the adverse effects of the attack. This large switching gain can lead to severe chattering, reducing the automatic steering performance and safety of the autonomous vehicle. Summary of the Invention

[0005] The purpose of this invention is to provide a safety control method, system, and device for automatic steering of unmanned vehicles.

[0006] The technical solution of this invention is as follows:

[0007] A safety control method for automatic steering of an autonomous vehicle includes the following operations:

[0008] S1 derives a dynamic state-space model based on the vehicle's kinematic and dynamic models.

[0009] S2 obtains the actual control signal under the actuator attack based on the dynamic state space model and the preset actuator attack signal. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal.

[0010] S3 selects the integral sliding surface based on the tracking error of the vehicle system to obtain the sliding mode approaching law; based on the sliding mode approaching law and the actual attack signal, the sliding mode controller signal is obtained;

[0011] Based on the integral sliding surface, S4 constructs a Lyapunov function, obtains the derivative of the Lyapunov function, and combines it with the sliding mode controller signal to obtain a sliding mode controller safety signal. The range of the sliding mode controller safety signal is controlled to not exceed a first threshold, thereby realizing the safety control of automatic steering of the unmanned vehicle.

[0012] According to the security control method of claim 1, the operation of obtaining the actual control signal under actuator attack in step S2 can be implemented by the following formula:

[0013]

[0014] The actual control signal under the actuator attack is u(t), and the sliding mode controller signal is u(t). The preset actuator attack signal is given, x(t) is the state variable in the dynamic state-space model, t is time, and ω is the actuator attack frequency.

[0015] In the safety control method described above, the tracking error of the vehicle system in S3 is the difference between the actual yaw rate and the ideal yaw rate of the vehicle.

[0016] The safety control method described above, wherein the operation of obtaining the sliding mode controller signal in S3 can be implemented by the following formula:

[0017]

[0018]

[0019] u(t) is the sliding mode controller signal. Here, x(t) is the actual attack signal, x(t) is the state variable in the dynamic state-space model, t is time, ω is the actuator attack frequency, κsgn(s(t)) is the sliding mode reaching law, κ is the gain of the sliding mode reaching law, and s(t) is the integral sliding surface; f C is the distance from the vehicle's center of gravity to the front axle. fFor the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r For the lateral stiffness of the rear wheels of the vehicle, I z Let v be the moment of inertia of the vehicle about the z-axis. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r This refers to the vehicle's actual yaw rate. Let λ be the derivative of the ideal yaw rate of the vehicle, λ be the sliding surface coefficient, and e be the tracking error of the vehicle system.

[0020] As described above, in the safety control method, the operation of controlling the range of the sliding mode controller safety signal to not exceed the first threshold in S4 specifically involves: controlling the gain range of the sliding mode reaching law to not exceed the second threshold, and designing an adaptive rate to eliminate redundant data in the sliding mode controller safety signal, thereby achieving the control of the range of the sliding mode controller safety signal to not exceed the first threshold.

[0021] The safety control method described above, specifically the operation of obtaining the derivative of the Lyapunov function in step S4, is as follows:

[0022] Based on the estimation bias of the integral sliding surface and network weights, a Lyapunov function is constructed. s(t) is the integral sliding surface. This refers to the estimation bias of the network weights. λ is the transpose of the estimation bias of the network weights, and λ is the sliding surface coefficient.

[0023] Differentiating the Lyapunov function yields its derivative.

[0024]

[0025]

[0026] s(t) is the integral sliding surface, l f C is the distance from the vehicle's center of gravity to the front axle. f For the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r For the lateral stiffness of the rear wheels of the vehicle, I z Let v be the moment of inertia of the vehicle about the z-axis. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r ω is the actual yaw rate of the vehicle. d For the ideal yaw rate of the vehicle, Let u(t) be the derivative of the ideal yaw rate of the vehicle, and u(t) be the sliding mode controller signal. The preset actuator attack signal is given, where x(t) is a state variable in the dynamic state-space model, t is time, and ω is the actuator attack frequency. λ is the derivative of the estimated network weights, λ is the sliding surface coefficient, and γ is the Lyapunov function model parameter.

[0027] As described above, the safety control method further includes, after the operation of controlling the range of the sliding mode controller safety signal to not exceed the threshold in step S4, obtaining the optimal steering wheel angle based on the condition that the range of the sliding mode controller safety signal does not exceed the threshold, and replacing the initial steering wheel angle with the optimal steering wheel angle to achieve safe control of automatic steering of the unmanned vehicle.

[0028] The optimal steering wheel angle can be obtained using the following formula:

[0029] δ sw =δ f *i sw ,

[0030]

[0031] δ sw δ is the optimal steering wheel angle. f i is the steering angle of the vehicle's front wheels. sw I is the transmission ratio between the steering wheel angle and the wheel angle of a vehicle. z Let l be the moment of inertia of the vehicle about the z-axis. f C is the distance from the vehicle's center of gravity to the front axle. f For the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r v is the lateral stiffness of the rear wheel of the vehicle. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r ω is the actual yaw rate of the vehicle. d Let λ be the ideal yaw rate of the vehicle, λ be the sliding surface coefficient, and e be the tracking error of the vehicle system. This is the preset executor attack signal. The actual attack signal is κsgn(s(t)), the sliding mode reaching law is κsgn(s(t)), x(t) is the state variable in the dynamic state-space model, t is time, and ω is the actuator attack frequency.

[0032] A safety control system for automatic steering of an autonomous vehicle includes:

[0033] The dynamic state-space model generation module is used to obtain a dynamic state-space model based on the vehicle's kinematic and dynamic models.

[0034] The actual attack signal generation module is used to obtain the actual control signal under the actuator attack based on the dynamic state space model and the preset actuator attack signal. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal.

[0035] The sliding mode controller signal generation module is used to select the integral sliding surface based on the tracking error of the vehicle system to obtain the sliding mode reaching law; and to obtain the sliding mode controller signal based on the sliding mode reaching law and the actual attack signal.

[0036] The sliding mode controller safety signal generation and range control module is used to construct a Lyapunov function based on the integral sliding surface, obtain the derivative of the Lyapunov function, combine it with the sliding mode controller signal to obtain the sliding mode controller safety signal, and control the range of the sliding mode controller safety signal to not exceed a first threshold, thereby realizing the safety control of automatic steering of the unmanned vehicle.

[0037] A safety control device for automatic steering of an unmanned vehicle includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the above-described safety control method for automatic steering of an unmanned vehicle.

[0038] A computer-readable storage medium is characterized in that it is used to store a computer program, wherein the computer program, when executed by a processor, implements the above-described safety control method for automatic steering of an unmanned vehicle.

[0039] The beneficial effects of this invention are as follows:

[0040] This invention provides a safety control method for automatic steering of unmanned vehicles. Addressing the phenomenon of actuator attacks during the automatic steering control process of unmanned vehicles, a preset actuator attack signal without specific conditions is designed to obtain the actual control signal input under actuator attack. After nonlinear mapping and weighted summation processing using a neural network method, an accurate actual attack signal is obtained. Then, using an integral sliding mode strategy, a sliding mode reaching law is designed and combined with the actual attack signal to construct a sliding mode controller signal. This signal is then combined with the derivative of a designed Lyapunov function and the range of the obtained sliding mode controller safety signal is controlled, enabling automatic steering of the unmanned vehicle and achieving safe control of its automatic steering.

[0041] This invention provides a safety control method for automatic steering of unmanned vehicles. The method assumes no special rules for the preset actuator attack signal and approximates the preset actuator attack signal through a neural network. The resulting actual attack signal is closer to the actual situation and has higher accuracy compared to network attack signals designed according to certain specific probability distribution conditions in the prior art.

[0042] This invention provides a safety control method for automatic steering of unmanned vehicles. It designs a safety control strategy based on integral sliding surface, which, while ensuring the steering performance of unmanned vehicles, utilizes the inherent robustness of sliding control to mitigate the adverse effects of actuator attacks, so that unmanned vehicles maintain good steering performance in actuator attack environments.

[0043] This invention provides a safety control method for automatic steering of unmanned vehicles. It uses a radial basis function neural network to approximate and reconstruct the preset actuator attack signal, and constructs a sliding mode controller signal by combining a sliding mode approaching law designed by a nonlinear control method. This not only alleviates the adverse effects of actuator attacks on the automatic steering control of unmanned vehicles, but also solves the problem of vehicle vibration. Attached Figure Description

[0044] The solutions and advantages of this application will become clear to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0045] In the attached diagram:

[0046] Figure 1 This is a flowchart illustrating the security control method in the embodiment;

[0047] Figure 2 This is a structural schematic diagram of the vehicle system execution flowchart in the embodiment;

[0048] Figure 3 This is a schematic diagram of the safety control system in the embodiment;

[0049] Figure 4 This is a schematic diagram of the safety control device in the embodiment. Detailed Implementation

[0050] Exemplary embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings.

[0051] This embodiment provides a safety control method for automatic steering of an autonomous vehicle. See [link to relevant documentation] Figure 1 This includes the following operations:

[0052] S1 derives a dynamic state-space model based on the vehicle's kinematic and dynamic models.

[0053] S2 obtains the actual control signal under the actuator attack based on the dynamic state space model and the preset actuator attack signal. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal.

[0054] S3 selects the integral sliding surface based on the tracking error of the vehicle system to obtain the sliding mode approaching law; based on the sliding mode approaching law and the actual attack signal, the sliding mode controller signal is obtained;

[0055] S4 constructs a Lyapunov function based on the integral sliding surface, obtains the derivative of the Lyapunov function, and combines it with the sliding mode controller signal to obtain a sliding mode controller safety signal. The range of the sliding mode controller safety signal is controlled to not exceed a first threshold, thereby achieving safe control of the automatic steering of the autonomous vehicle. S1 obtains a dynamic state-space model based on the vehicle's kinematic and dynamic models.

[0056] A kinematic and dynamic model of the vehicle is constructed. The kinematic and dynamic models are combined to obtain a two-degree-of-freedom dynamic model of the vehicle's automatic steering control model. Then, the two-degree-of-freedom dynamic model of the vehicle's automatic steering control model is simplified and transformed into a dynamic state-space model with respect to longitudinal velocity and yaw rate.

[0057] Construct a kinematic model of the vehicle. Considering a two-degree-of-freedom vehicle model, neglecting the effects of the vehicle suspension and road slope on the vehicle, applying Newton's second law of kinematics, we obtain:

[0058] m(v y +v x ω r ) = F yf +F yr ,

[0059] In the formula, m is the total vehicle mass, and v x v is the lateral velocity of the vehicle. y Let ω be the lateral velocity of the vehicle. r F is the yaw rate of the vehicle. yf F yr These are the lateral forces of the front and rear tires of the vehicle, respectively.

[0060] Construct a dynamic model of the vehicle. By establishing torque balance around the z-axis, the yaw dynamic equations are obtained as follows:

[0061]

[0062] In the formula, l f and l r These are the distances from the vehicle's center of gravity to the front and rear axles, respectively.

[0063] The lateral forces on the front and rear wheels of the vehicle are respectively: F yf =2C f (δ f -θ Vf ), F yr =2C r (-θVr ), C f C represents the lateral stiffness of the front wheel. r For the lateral stiffness of the rear wheel, θ Vf θ Vr These are the front and rear wheel speed angles of the vehicle, δ f This refers to the steering angle of the vehicle's front wheels.

[0064] Due to θ Vf θ Vr Comparing the smaller values, we get:

[0065]

[0066] Combining the kinematic and dynamic models of the vehicle described above, the resulting two-degree-of-freedom dynamic model is as follows:

[0067]

[0068]

[0069] I z Let be the moment of inertia of the vehicle about the z-axis.

[0070] Simplifying the above equation, we obtain the following dynamic state-space model of the vehicle with respect to its longitudinal velocity and yaw rate:

[0071]

[0072] y(t)=Cx(t),

[0073] Regarding the state-space model of the dynamics of an autonomous vehicle system, the state variable x(t) = [v y ω r ] T The control quantity u(t) = δ f ,

[0074]

[0075]

[0076] C is a 1×2 identity matrix.

[0077] S2 obtains the actual control signal under actuator attack based on the dynamic state-space model and the preset actuator attack signal. The actual control signal under actuator attack is then processed by nonlinear mapping and weighted summation to obtain the actual attack signal.

[0078] Figure 2 The vehicle system execution flowchart, from Figure 2As can be seen, the automatic steering control actions of autonomous vehicles are generated by actuators, and their performance may be affected by malicious attacks or external disturbances. Previous studies have shown that most cyberattack signals are assumed to follow a specific probability distribution. However, actual cyberattacks do not follow a specific rule, and their information is difficult to obtain. Therefore, the cyberattack signal information obtained by existing technologies differs significantly from the actual situation. Thus, this embodiment defines a preset actuator attack signal without any special assumptions as... The actual control signal under an actuator attack can be calculated using the following formula:

[0079]

[0080] The actual control signal under actuator attack, u(t) is the sliding mode controller signal. Here, x(t) represents the preset actuator attack signal, x(t) represents the state variable in the dynamic state-space model, t represents time, and ω represents the actuator attack frequency.

[0081] Combining the actual control signal under actuator attack with the above-mentioned dynamic state-space model, the state-space expression of the autonomous vehicle's automatic steering control system can be obtained as follows:

[0082]

[0083] Where K is the gain of the state feedback controller.

[0084] To improve the stability of the automatic steering system of autonomous vehicles under actuator attacks and to address the chattering issue inherent in traditional sliding mode control, this embodiment processes the actual control signal under actuator attacks through nonlinear mapping and weighted summation. Specifically, a radial basis function (RBF) neural network is used to approximate the actual control signal under actuator attacks, forcing a preset actuator attack signal to display its data information, thus obtaining the actual attack signal.

[0085] The RBF network algorithm is as follows:

[0086]

[0087]

[0088] Where x is the network input, μ is the center vector of the Gaussian function, η is the base width of the Gaussian function, i is the number of network inputs, j is the j-th node of the hidden layer of the network, and φ = [φ1, φ2, ..., φ]. n ] T W is the output of the Gaussian function. * W represents the ideal value for the network weights.*T Let ε be the ideal transpose of the network weights, and ε be the network's approximation error, |ε|≤ε N .

[0089] Using RBF neural networks to approximate unknown preset actuator attack signals The network input is x = [v y ω r ] T The actual attack signal output by the RBF neural network is:

[0090]

[0091] but

[0092]

[0093] in W *T The ideal values ​​of the network weights are transposed. These are estimates of the network weights. The estimated values ​​of the network weights are transposed. It is the estimation bias of network weights. This is the transpose of the estimated network weights.

[0094] RBF neural networks are characterized by their ability to approximate arbitrary nonlinear functions with arbitrary precision. In an RBF neural network, the input layer receives the state variables from the dynamic state-space model, the intermediate layers consist of a set of radial basis functions (RBFs), and the output layer performs a weighted summation of the RBF output values ​​to obtain the final actual attack signal. The RBFs are typically Gaussian or polynomial functions used to nonlinearly map the state variables in the dynamic state-space model, transforming them from the original state space to a higher-dimensional feature space, thereby expanding the model's expressive power. Training an RBF neural network typically employs forward propagation and backpropagation algorithms. Forward propagation calculates the network's output, while backpropagation calculates the error and updates the network parameters, enabling the network to progressively approximate the target function.

[0095] S3 selects the integral sliding surface based on the tracking error of the vehicle system and obtains the sliding mode reaching law; based on the sliding mode reaching law and the actual attack signal, the sliding mode controller signal is obtained.

[0096] The tracking error of the vehicle system is the actual yaw rate ω of the vehicle. r With ideal yaw rate ω d The difference, i.e., e = ω r -ω d Select the integral sliding surface λ is the sliding surface coefficient, λ > 0; based on the integral sliding surface, a nonlinear control method is used to design the sliding mode reaching law. κ is the gain of the sliding mode reaching law.

[0097] Based on sliding mode convergence law and actual attack signals The sliding mode controller signal can be calculated using the following formula:

[0098]

[0099]

[0100] u(t) is the sliding mode controller signal. The actual attack signal is given by x(t), which is the state variable in the dynamic state-space model, t is time, ω is the actuator attack frequency, κsgn(s(t)) is the sliding mode reaching law, κ is the gain of the sliding mode reaching law, and s(t) is the integral sliding surface; f C is the distance from the vehicle's center of gravity to the front axle. f For the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r For the lateral stiffness of the rear wheels of the vehicle, I z Let v be the moment of inertia of the vehicle about the z-axis. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r This refers to the vehicle's actual yaw rate. Let λ be the derivative of the ideal yaw rate of the vehicle, λ be the sliding surface coefficient, and e be the tracking error of the vehicle system.

[0101] S4 constructs a Lyapunov function based on an integral sliding surface, obtains the derivative of the Lyapunov function, and combines it with the sliding mode controller signal to obtain the sliding mode controller safety signal. The range of the sliding mode controller safety signal is controlled to not exceed a first threshold, thereby realizing the safe control of automatic steering of the unmanned vehicle.

[0102] The specific steps to obtain the derivative of the Lyapunov function are as follows: based on the integral sliding surface and the estimation bias of the network weights, construct the Lyapunov function. To obtain the derivative of the Lyapunov function by taking the transpose of the estimation bias of the network weights:

[0103]

[0104]

[0105] Let s(t) be the derivative of the estimated network weights, and l be the integral sliding surface.f C is the distance from the vehicle's center of gravity to the front axle. f For the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r For the lateral stiffness of the rear wheels of the vehicle, I z Let v be the moment of inertia of the vehicle about the z-axis. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r ω is the actual yaw rate of the vehicle. d For the ideal yaw rate of the vehicle, Let be the derivative of the ideal yaw rate of the vehicle, e be the tracking error of the vehicle system, and u(t) be the sliding mode controller signal. Here, x(t) represents the preset actuator attack signal, x(t) is the state variable in the dynamic state-space model, t is time, and ω is the actuator attack frequency. λ is the derivative of the estimated network weights, λ is the sliding surface coefficient, γ is the Lyapunov function model parameter, and γ is the model parameter related to the Lyapunov function.

[0106] Substituting the sliding mode controller signal into the derivative of the Lyapunov function, the sliding mode controller safety signal is obtained as follows:

[0107]

[0108] When the range of the sliding mode safety control signal does not exceed a first threshold, redundant data in the sliding mode controller's safety signal is eliminated by controlling the gain range of the sliding mode reaching law to not exceed a second threshold and designing an adaptive rate. Specifically, the gain κ of the sliding mode reaching law is set to ≥ |B 12 ε| max And design adaptive rate and Equal, thus stabilizing the sliding mode safety control signal. This ensures that the automatic steering safety controller for autonomous vehicles meets system stability requirements.

[0109] After controlling the range of the sliding mode controller's safety signal to not exceed the threshold, the process also includes obtaining the optimal steering wheel angle, which replaces the initial steering wheel angle to achieve safe control of the autonomous vehicle's automatic steering.

[0110] make The steering angle δ of the vehicle's front wheels at this moment is obtained. f :

[0111]

[0112] This leads to the optimal steering wheel angle δ. sw =δ f *i sw isw δ is the transmission ratio between the steering wheel angle and the wheel angle. sw For the optimal steering wheel angle, I z Let l be the moment of inertia of the vehicle about the z-axis. f C is the distance from the vehicle's center of gravity to the front axle. f For the lateral stiffness of the vehicle's front wheels, l r C is the distance from the vehicle's center of gravity to the rear axle. r v is the lateral stiffness of the rear wheel of the vehicle. y v is the lateral velocity of the vehicle. x Let ω be the longitudinal velocity of the vehicle. r ω is the actual yaw rate of the vehicle. d Let λ be the ideal yaw rate of the vehicle, λ be the sliding surface coefficient, and e be the tracking error of the vehicle system. Preset executor attack signal, κsgn(s(t)) represents the actual attack signal, κsgn(s(t)) represents the sliding mode reaching law, x(t) represents the state variable in the dynamic state-space model, t represents time, and ω represents the actuator attack frequency.

[0113] This embodiment provides a safety control system for automatic steering in an autonomous vehicle. See [link / reference] Figure 3 ,include:

[0114] The dynamic state-space model generation module is used to obtain a dynamic state-space model based on the vehicle's kinematic and dynamic models.

[0115] The actual attack signal generation module is used to obtain the actual control signal under the actuator attack based on the dynamic state space model and the preset actuator attack signal. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal.

[0116] The sliding mode controller signal generation module is used to select the integral sliding surface based on the tracking error of the vehicle system and obtain the sliding mode reaching law; based on the sliding mode reaching law and the actual attack signal, the sliding mode controller signal is obtained.

[0117] The sliding mode controller safety signal generation and range control module is used to construct a Lyapunov function based on the integral sliding surface, obtain the derivative of the Lyapunov function, and combine it with the sliding mode controller signal to obtain the sliding mode controller safety signal. The module controls the range of the sliding mode controller safety signal to not exceed a first threshold, thereby realizing the safety control of automatic steering of the unmanned vehicle.

[0118] This embodiment provides a safety control device for automatic steering of an unmanned vehicle. See [link / reference] Figure 4 It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the aforementioned safety control method for automatic steering of unmanned vehicles.

[0119] This embodiment provides a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned safety control method for automatic steering of an unmanned vehicle.

[0120] This embodiment provides a safety control method for automatic steering of unmanned vehicles. Addressing the phenomenon of actuator attacks during the automatic steering control process of unmanned vehicles, a preset actuator attack signal without specific assumptions is designed to obtain the actual control signal input under actuator attack. After nonlinear mapping and weighted summation processing in a neural network method, an accurate actual attack signal is obtained. Then, using an integral sliding mode strategy, a sliding mode reaching law is designed and combined with the actual attack signal to construct a sliding mode controller signal. This signal is then combined with the derivative of a designed Lyapunov function and the range of the obtained sliding mode controller safety signal is controlled, enabling automatic steering of the unmanned vehicle and achieving safe control of its automatic steering.

[0121] This embodiment provides a safety control method for automatic steering of unmanned vehicles. The preset actuator attack signal has no special assumptions or rules, and the preset actuator attack signal is approximated by a neural network. The actual attack signal obtained is closer to the actual situation and has higher accuracy than the network attack signal designed according to a certain probability distribution condition in the prior art.

[0122] This embodiment provides a safety control method for automatic steering of unmanned vehicles. It designs a safety control strategy based on integral sliding surface, which, while ensuring the steering performance of unmanned vehicles, utilizes the inherent robustness of sliding control to mitigate the adverse effects of actuator attacks, enabling unmanned vehicles to maintain good steering performance in actuator attack environments.

[0123] This embodiment provides a safety control method for automatic steering of unmanned vehicles. It uses a radial basis function neural network to approximate and reconstruct the preset actuator attack signal, and constructs a sliding mode controller signal by combining a sliding mode approaching law designed by a nonlinear control method. This not only alleviates the adverse effects of actuator attacks on the automatic steering control of unmanned vehicles, but also solves the problem of vehicle vibration.

Claims

1. A safety control method for automatic steering of an unmanned vehicle, characterized in that, This includes the following operations: S1 derives a dynamic state-space model based on the vehicle's kinematic and dynamic models; S2 Based on the dynamic state-space model and the preset actuator attack signal, the actual control signal under the actuator attack is obtained. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal. S3. Based on the tracking error of the vehicle system, an integral sliding surface is selected to obtain the sliding mode reaching law; based on the sliding mode reaching law and the actual attack signal, the sliding mode controller signal is obtained, and the calculation formula is as follows: ， ， ， For sliding mode controller signals, For the frequency of executor attacks, The derivative of the ideal yaw rate of the vehicle. The sliding surface coefficient, For the tracking error of the vehicle system, This is an actual attack signal. These are the state variables in the dynamic state-space model. t For time, For sliding mode reaching law, For the gain of the sliding mode reaching law, For integral sliding surfaces; The distance from the vehicle's center of gravity to the front axle. This refers to the lateral stiffness of the vehicle's front wheels. This is the distance from the vehicle's center of gravity to the rear axle. For the lateral stiffness of the vehicle's rear wheels, For vehicles to bypass Moment of inertia of the shaft For the vehicle's lateral speed, For the longitudinal speed of the vehicle, This refers to the vehicle's actual yaw rate. S4 Based on the integral sliding surface, a Lyapunov function is constructed, the derivative of the Lyapunov function is obtained, and combined with the sliding mode controller signal, a sliding mode controller safety signal is obtained. The range of the sliding mode controller safety signal is controlled to not exceed a first threshold, thereby realizing the safety control of automatic steering of the unmanned vehicle.

2. The safety control method according to claim 1, characterized in that, The operation in S2 to obtain the actual control signal under actuator attack can be achieved by the following formula: ， The actual control signal under actuator attack. This is a preset executor attack signal.

3. The safety control method according to claim 1, characterized in that, The tracking error of the vehicle system in S3 is the difference between the actual yaw rate and the ideal yaw rate of the vehicle.

4. The safety control method according to claim 1, characterized in that, The operation in S4 that controls the range of the sliding mode controller's safety signal to not exceed the first threshold specifically involves: By controlling the gain range of the sliding mode reaching law to not exceed a second threshold, and designing an adaptive rate to eliminate redundant data in the sliding mode controller's safety signal, the range of the sliding mode controller's safety signal can be controlled to not exceed a first threshold.

5. The safety control method according to claim 1, characterized in that, The specific operation for obtaining the derivative of the Lyapunov function in S4 is as follows: Based on the estimation bias of the integral sliding surface and network weights, a Lyapunov function is constructed. , This refers to the estimation bias of the network weights. This is the transpose of the estimation bias of the network weights; Differentiating the Lyapunov function yields its derivative. ， For the ideal yaw rate of the vehicle, For preset executor attack signals, The derivative of the estimated network weights. These are the parameters of the Lyapunov function model.

6. The safety control method according to claim 1, characterized in that, After the operation in S4 that controls the range of the sliding mode controller's safety signal to not exceed the threshold, the following is also included: Based on the condition that the range of the sliding mode controller safety signal does not exceed the threshold, the optimal steering wheel angle is obtained. The optimal steering wheel angle replaces the initial steering wheel angle to achieve safe control of automatic steering of the unmanned vehicle. The optimal steering wheel angle can be obtained using the following formula: ， ； To achieve the optimal steering wheel angle, This refers to the steering angle of the vehicle's front wheels. This is the transmission ratio between the steering wheel angle and the wheel angle. For the ideal yaw rate of the vehicle, This is a preset executor attack signal.

7. A safety control system for automatic steering of an unmanned vehicle, used to implement the safety control method of claim 1, characterized in that, include: The dynamic state-space model generation module is used to obtain a dynamic state-space model based on the vehicle's kinematic and dynamic models. The actual attack signal generation module is used to obtain the actual control signal under the actuator attack based on the dynamic state space model and the preset actuator attack signal. The actual control signal under the actuator attack is processed by nonlinear mapping and weighted summation to obtain the actual attack signal. The sliding mode controller signal generation module is used to select the integral sliding surface based on the tracking error of the vehicle system and obtain the sliding mode reaching law; Based on the sliding mode approach law and the actual attack signal, the sliding mode controller signal is obtained; The sliding mode controller safety signal generation and range control module is used to construct a Lyapunov function based on the integral sliding surface, obtain the derivative of the Lyapunov function, combine it with the sliding mode controller signal to obtain the sliding mode controller safety signal, and control the range of the sliding mode controller safety signal to not exceed a first threshold, thereby realizing the safety control of automatic steering of the unmanned vehicle.

8. A safety control device for automatic steering of an unmanned vehicle, characterized in that, It includes a processor and a memory, wherein the processor executes a computer program stored in the memory to implement the safety control method for automatic steering of an unmanned vehicle as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the safety control method for automatic steering of an unmanned vehicle as described in any one of claims 1-6.

Images