Intelligent vehicle lateral control method based on sensor measurement reconstruction under network attack

Abstract

The application relates to a lateral control method for intelligent vehicles based on sensor measurement reconstruction under network attacks, and relates to intelligent safety and automatic driving of vehicles. A model predictive control (MPC) controller based on sensor measurement reconstruction is constructed to realize accurate control of the lateral movement of a vehicle. The system monitors sensor data in real time, detects abnormal sensor data caused by network attacks, and starts a measurement reconstruction mechanism to correct the data by using redundant sensors or algorithm models. A lower torque optimization and distribution controller is designed, which intelligently adjusts the output torque of four wheel hub motors according to the expected yaw moment output by the upper MPC controller and the state information of the vehicle, so as to realize accurate control of the yaw moment of the vehicle, reduce the lateral tracking error, and improve the driving stability and dynamic performance of the vehicle. A lateral cooperative control rule is proposed to coordinate the work of the upper MPC controller and the lower torque optimization and distribution controller, so that the overall optimization of the lateral control of the vehicle is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the intelligent safety and automatic driving of an automobile, and in particular to an intelligent automobile lateral control method based on sensor measurement reconstruction under network attacks.

[0002] BACKGROUND

[0003] With the development of social economy, intelligent electric vehicles have been rapidly developed and widely used in various fields. Control is an important link that affects the motion state of vehicles, among which trajectory tracking control and lateral stability control are closely related to the safety of autonomous driving. Current vehicle control technology can meet the automatic driving control of vehicles under the condition that the information of the vehicle itself and the external environment is true and accurate. However, in the environment of intelligent networking, the sensing information of the vehicle is easy to be attacked by the network, and malicious attacks make the sensing data and lateral control input signals at the information level untrustworthy, resulting in phenomena such as steering wheel failure and loss of control of autonomous vehicles. Under the untrustworthy state input of the vehicle, the commonly used state estimation methods such as Kalman filter are used to correct the attacked signals, so as to ensure the normal function of the controller. Literature 1 (Zhang J, Zhang B, Zhang N, et al. A novel robust event-triggered fault tolerant automatic steering control approach of autonomous land vehicles under in-vehicle network delay[J]. International Journal of Robust and Nonlinear Control, 2021, 31(7): 2436-2464.) established an event-triggered automatic fault-tolerant steering control system for vehicle network delay, which can significantly improve the resource utilization rate of the limited bandwidth of the vehicle network and ensure the asymptotic stability of the steering control system. Literature 2 (Guo J, Wang J, Luo Y, et al. Takagi-Sugeno fuzzy-based robust H infinity integrated lane-keeping and direct yaw moment controller of unmanned electric vehicles[J]. IEEE / ASME Transactions on Mechatronics, 2021, 26(4):2151-2162.) designed a steering control multi-model set representing the time-varying of vehicle parameters, time delay, parameter uncertainty and input saturation characteristics, and used T-S fuzzy control theory to design the lateral motion controller of autonomous vehicles.

[0004] In addition to precise path tracking capability, vehicles need to have strong robustness and fault tolerance to malicious intrusion under network attacks. In the existing research, although the lateral motion control of autonomous vehicles has been studied in depth, the influence of malicious attacks on vehicles is not considered in current research, which may lead to controller failure and complete deviation of vehicle travel from the desired path, causing serious traffic accidents. SUMMARY

[0005] The purpose of the present application is to solve the control problem of intelligent electric vehicles disturbed by network attacks when obtaining their own motion state, and to provide an intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks, which can enable intelligent electric vehicles to still input accurate state estimation information to the model predictive control (MPC) controller when the sensors are attacked by network attacks, and ensure the lateral stability and safety of the vehicle.

[0006] The technical solution of the present application is to collect the driving state information of intelligent electric vehicles through the vehicle-mounted sensor system, consider that the sensors are attacked by network attacks, and design an autonomous vehicle lateral safety control strategy based on sensor measurement reconstruction based on the existing vehicle state estimation and attack signal estimation. The present application comprises the following steps:

[0007] Step 1: A three-degree-of-freedom dynamic model describing the lateral, longitudinal and yaw dynamics of intelligent electric vehicles is established, the conversion relationship of the longitudinal and lateral forces received by the tires in the x-axis and y-axis directions of the vehicle body coordinate system is simplified to obtain a nonlinear state space equation, and the equation is discretized, the process noise and measurement noise in state estimation are introduced, and the influence of network attack signals and data packet loss on sensor measurement output signals is added;

[0008] 1.1) A three-degree-of-freedom dynamic model accurately representing the behavior mechanism of intelligent electric vehicles is established, which includes the lateral, longitudinal and yaw dynamics of the vehicle;

[0009] 1.2) The conversion relationship of the longitudinal and lateral forces received by the tires in the x-axis and y-axis directions of the vehicle body coordinate system is established;

[0010] 1.3) The nonlinear state space equation of the intelligent electric vehicle is simplified;

[0011] 1.4) The first-order difference quotient method is used to approximate and discretize the above nonlinear state space equation to obtain a nonlinear discrete state equation;

[0012] 1.5) Process noise and measurement noise are introduced to reflect the uncertainty in the actual driving process, and the state equation and measurement equation of state estimation are established based on the nonlinear discrete state equation;

[0013] 1.6) Introduce network attack signal, establish measurement equation of sensor suffering network attack;

[0014] 1.7) Introduce data random packet loss, establish measurement equation of sensor data packet loss;

[0015] Step 2: Design an upper MPC controller based on sensor measurement reconstruction; when the vehicle-mounted sensor system is subjected to malicious network attacks, the attack detection result and attack signal estimation result are used for real-time switching of the control law, so as to reduce the influence of network attacks on the lateral control system performance of the autonomous vehicle;

[0016] 2.1) Assume that the vehicle state information obtained through the vehicle-mounted sensor system has been disturbed by network attacks and packet loss, and accurate state estimation is obtained through state estimation method;

[0017] 2.2) Match the lateral control system model affected by network attacks with the nominal lateral control system model;

[0018] 2.3) Introduce state estimation disturbed by network attacks and packet loss, and reconstruct the output state prediction model in the upper MPC controller;

[0019] 2.4) Set constraint conditions for the state quantity, control quantity and control increment of the vehicle;

[0020] 2.5) Set the objective function for the reconstructed upper MPC controller and solve it to obtain the optimal control strategy;

[0021] Step 3: Design a lower torque optimization and distribution controller; establish a torque optimization and distribution method to realize the control of vehicle yaw moment;

[0022] 3.1) Design the objective function with the minimum of the comprehensive utilization rate of the four wheels as the optimization target;

[0023] 3.2) Set constraint conditions according to road adhesion conditions and hub motor performance;

[0024] 3.3) Convert the optimization problem into a quadratic programming problem to solve the torque of the four wheels;

[0025] Step 4: Construct a lateral cooperative control rule, consider the network attack factor, use the surplus yaw stability of the vehicle to reduce the lateral tracking error, and realize the cooperative work between the upper MPC controller and the lower torque optimization and distribution controller through the construction of the lateral cooperative control rule, to jointly cope with the influence of network attacks on the lateral control of the vehicle.

[0026] Compared with the prior art, the present application has the following outstanding technical effects and advantages:

[0027] 1、The application can provide accurate state estimation information for the MPC controller of the intelligent electric vehicle under network attacks by introducing a lateral control method based on sensor measurement reconstruction. The problem of sensor data pollution caused by network attacks is effectively avoided, thereby significantly improving the network security protection capability of the vehicle in the intelligent network environment.

[0028] 2、The application designs an MPC controller based on sensor measurement reconstruction, which can detect and respond to the influence of network attacks on sensor data in real time. By dynamically adjusting the control strategy, the controller can ensure that the vehicle maintains stable lateral control performance when attacked, avoiding safety problems such as steering wheel failure and loss of control.

[0029] 3、The application realizes accurate control of vehicle yaw moment by constructing a lower torque optimization distribution controller. Not only helps to reduce lateral tracking error, but also can intelligently adjust according to road adhesion conditions and hub motor performance, thereby optimizing the driving stability and dynamics performance of the vehicle.

[0030] 4、The application proposes a lateral cooperative control rule, which coordinates the work of the upper MPC controller and the lower torque optimization distribution controller to realize comprehensive optimization of vehicle lateral control. This cooperative control strategy not only improves the lateral tracking accuracy of the vehicle, but also enhances the overall efficiency and reliability of the system.

[0031] 5、The technical solution of the application is not only suitable for intelligent electric vehicles, but also can be extended to other types of intelligent network vehicles. In addition, the technology can be flexibly adjusted according to different vehicle models and different driving environments, and has wide adaptability and application prospect. It improves the safety and reliability of intelligent electric vehicles and lays a solid foundation for the widespread application of automatic driving technology. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 The figure is a schematic diagram of the lateral safety control strategy of the autonomous vehicle based on sensor measurement output reconstruction.

[0033] Figure 2 The figure is a schematic diagram of the control law reconstruction based on the model matching idea.

[0034] Figure 3 The figure is a schematic diagram of the lateral cooperative control rule.

[0035] Figure 4 The figure is a simulation verification of the intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks. (a) is the trajectory comparison; (b) is the yaw angle comparison; (c) is the yaw angular velocity comparison; (d) is the lateral acceleration comparison; (e) is the front wheel steering angle comparison; (f) is the wheel torque. DETAILED DESCRIPTION

[0036] To make the technical content, structural features, and implementation objectives of the present application clear, the present application is explained in detail below with reference to the accompanying drawings.

[0037] The general working idea of the present application is that when the intelligent electric vehicle suffers a network attack during normal driving, the network attack detection module is used for network attack detection, when the sensor is detected to be attacked by the network, the sensor attacked by the network is shielded immediately and the state estimation is performed by using the sensor information not attacked by the network, and the accurate estimation information is input into the controller to ensure the lateral safety control of the vehicle. Figure 1 As shown in the figure, the embodiment of the present application mainly includes the following steps:

[0038] Step 1: Establishing a dynamics model of an intelligent electric vehicle under network attack.

[0039] Step 1.1, according to Newton's second law, the force balance equations of longitudinal, lateral and yaw and the conversion equations of the vehicle body coordinate system and the inertial coordinate system are obtained:

[0040] (1)

[0041] In the formula, , are the distances from the mass center to the front and rear axles, is the vehicle curb weight, is the moment of inertia of the vehicle around the z axis, is the lateral velocity of the vehicle, is the longitudinal velocity of the vehicle, is the lateral acceleration of the vehicle, is the longitudinal acceleration of the vehicle, is the yaw angle of the vehicle, is the yaw angular velocity of the vehicle, is the lateral velocity in the earth coordinate system, is the longitudinal velocity in the earth coordinate system. is the force in the x axis direction of the front wheel, is the force in the x axis direction of the rear wheel, is the force in the y axis direction of the front wheel, is the force in the y axis direction of the rear wheel.

[0042] Step 1.2, the conversion relationship between the combined forces in the x axis and y axis directions of the tire in the vehicle body coordinate system and the longitudinal and lateral forces is as follows:

[0043] (2)

[0044] In the formula, is the force in the x axis direction of the front wheel, is the force in the x axis direction of the rear wheel, Fyf is the force in y-axis direction of front wheels, Fyf is the force in y-axis direction of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels.

[0045] Assuming that the intelligent electric vehicle tire cornering characteristics are in the linear range, the longitudinal force and lateral force of the tire are obtained by the following formulas:

[0046] (3)

[0047] In the formula, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels, Fxf is the longitudinal force of front wheels. Kxf is the longitudinal stiffness of front tire; Kxf is the longitudinal stiffness of front tire; Kxf is the longitudinal stiffness of front tire; Kxf is the longitudinal stiffness of front tire; μxf is the slip ratio of front tire; μxf is the slip ratio of front tire; αxf is the cornering angle of front tire; αxf is the cornering angle of front tire;

[0048] By simplifying the calculation, the calculation relationship of the tire cornering angle can be obtained:

[0049] (4)

[0050] In the formula, αxf is the cornering angle of front tire, αxf is the cornering angle of front tire, v is the lateral velocity of vehicle, ω is the yaw rate of vehicle, v is the longitudinal velocity of vehicle; , L is the distance from the center of mass to the front and rear axles, respectively; δf is the front wheel deflection angle.

[0051] Step 1.3, by solving formulas (1)-(4) simultaneously, the nonlinear intelligent electric vehicle dynamics model differential equation set is as follows:

[0052] (5)

[0053] In the formula, m is the vehicle kerb mass, This refers to the vehicle's longitudinal acceleration; The lateral acceleration of the vehicle. Let yaw acceleration be the acceleration of the vehicle. Let be the lateral speed of the vehicle. Let yaw rate be the vehicle's angular velocity. For the longitudinal stiffness of the front tire; For the longitudinal stiffness of the rear tire, For the lateral stiffness of the front tire, For the lateral stiffness of the rear tire, The front tire slip ratio, Rear tire slip ratio, Let be the moment of inertia of the vehicle about the z-axis. Let this be the vehicle's longitudinal velocity in the geodetic coordinate system. Let be the lateral velocity of the vehicle in the geodetic coordinate system.

[0054] Let the state variable be The control quantity is The nonlinear dynamic model represented by formula (5) can be abstracted into a nonlinear state-space equation, as shown in formula (6):

[0055] (6)

[0056] in, Represents the state variable at time t. Represents the control variable at time t. This represents the output variable at time t. This represents the first differential of the state variable at time t. Represents the abstract obtained about and The function, express and The function.

[0057] Step 1.4: Approximate discretization of the above nonlinear state-space equations using the first-order difference quotient method yields the nonlinear discrete state equations:

[0058] (7)

[0059] In the formula, for The state quantity at any given time. for The state quantity at any given time. Let k be the control variable at time k. The sampling period.

[0060] Step 1.5. Considering process noise and measurement noise, the above nonlinear discrete state equation and measurement equation of the autonomous vehicle can be integrated into the following form:

[0061] (8)

[0062] where, , represents the process noise, represents the measurement noise.

[0063] Step 1.6. Establish the sensor measurement equation that accurately characterizes the autonomous vehicle under network attack. The equation of the sensor data of the autonomous vehicle at time k under network attack can be described as:

[0064] (9)

[0065] where, is the output variable, is the malicious data injected by the attacker, is the sensor attack matrix set by the attacker.

[0066] Solve formula (8), (9) to establish the nonlinear discrete state equation and measurement equation of the vehicle under network attack:

[0067] (10)

[0068] where, represents the process noise, represents the measurement noise, is the output variable, is the malicious data injected by the attacker, is the sensor attack matrix set by the attacker. represents the state quantity at time k, represents the state quantity at time k.

[0069] Step 1.7. On the basis of formula (10), considering random packet loss, the nonlinear discrete state equation and measurement equation of the autonomous vehicle are as follows:

[0070] (11)

[0071] where, , , is the sensor measurement output, is an independent random variable introduced to describe the random measurement time delay and random packet loss phenomenon of the system. When the variable represents that the data is successfully transmitted through the in-vehicle communication network, no packet loss occurs, and the system is in a closed-loop state; when the variable , it represents that the data is not successfully transmitted through the in-vehicle communication network, packet loss occurs, and the system is in an open-loop state. The state transition probability matrix of the two-state Markov chain can be represented as:

[0072] (12)

[0073] where p ij represents the Markov chain state transition probability, is an independent random variable at time , and is an independent random variable at time .

[0074] Step 2: Designing an upper MPC controller based on sensor measurement reconstruction

[0075] Step 2.1, assuming that the vehicle state information obtained through the in-vehicle sensor system has been disturbed by network attacks and packet loss, while accurate state estimates are obtained through state estimation methods , and the attack signal is obtained through an attack signal estimator

[0076] (13)

[0077] where and are the attack vectors at time and time estimated by the attack signal estimator , which contain attack signals and communication packet loss disturbances, is the control input at time , and is the sensor measurement output at time .

[0078] Step 2.2, based on the model matching method, the lateral control system model affected by network attacks is matched with the nominal lateral control system model, as shown in Figure 2 . In the figure, is the reference trajectory of the lateral control system, is the control input, is the output of the vehicle model (10), is the output of the vehicle model (11), ​​represents the network attack disturbance to the sensor system. Combining the model matching idea, a new sensor measurement output equation is established based on the vehicle model (11) considering the non-ideal vehicle network environment and network attack disturbance:

[0079] (14)

[0080] where, represents the sensor measurement output injected by the network attack under the non-ideal communication condition, represents the reconstructed sensor measurement output, is the attack signal, is the network attack disturbance estimation.

[0081] Step 2.3, establish the output state prediction model of the controller.

[0082] First step, linearize the discrete state equation, where, and represent the state vector and control vector of the nonlinear system respectively, based on the working point at time , is the state vector at time , is the state vector at the th step of prediction, is the state vector at the th step, is the state vector at the th step after applying the control vector at the th step, represents the abstracted function about and , the expression is as follows:

[0083] (15)

[0084] Second step, perform first-order Taylor expansion at point , where, and represent the state vector and control vector of the approximate linear system respectively, to obtain the approximate linearized discrete state space equation:

[0085] (16)

[0086] where, is the system matrix based on the th step of prediction after linearization of the system at time ;

[0087] is the system matrix based on the At any moment, the first The control matrix for step prediction;

[0088] , for the system after approximate linearization based on At any moment, the first The state deviation predicted in the step.

[0089] The third step is to represent the output vector as follows:

[0090] (17)

[0091] In the formula, This indicates the output yaw angle and lateral position.

[0092] The fourth step is to estimate the state under network attacks and packet loss interference based on the above linear discrete state equations. Based on this, combined control quantity The controller state variables reconstructed from the measured input are obtained as follows:

[0093] (18)

[0094] Based on this state variable, establish a new state-space expression:

[0095] (19)

[0096] In the formula, The system matrix of the new system equations, To control the number of variables, The number of state variables. For dimension unit array; The control matrix for the new system equations; This is the output matrix of the new system equations; This refers to the state deviation introduced by the new system equations after approximate linearization; For the new system in the first The control increment applied step by step.

[0097] Step 5: Set the prediction step size and control step size of the reconstructed MPC controller to... and The control increment is only applied to the first... To the Step, that is:

[0098] (20)

[0099] Meanwhile, to simplify calculations, it is assumed that the system matrix and control matrix of the new system equations are in the _____. Step to the first The prediction in each step remains unchanged, and is the same as in the first step; since the output variables in each prediction step are the yaw angle and lateral position, the output matrix also remains unchanged; that is:

[0100] (21)

[0101] In the sixth step, based on the above assumptions, the reconstructed output (14) and the estimated attack vector (13) are combined to obtain the system prediction output expression based on the sensor measurement output reconstruction:

[0102] (22)

[0103] where, is the system prediction output in the prediction step;

[0104] is the sensor measurement output in the prediction step;

[0105] is the estimated attack vector in the prediction step;

[0106] is the control increment in the control step;

[0107] is the state deviation of the system approximation linearization in the prediction step;

[0108] is the state variable of the system at time

[0109] is the coefficient matrix of the system prediction output in the prediction step with respect to the system state variable known at time

[0110] is the coefficient matrix of the system prediction output in the prediction step with respect to the control increment in the control step;

[0111] is the coefficient matrix of the system prediction output in the prediction step with respect to the state deviation of the system approximation linearization;

[0112] , is the attack signal selection matrix in the prediction step.

[0113] Step 2.4, considering that the autonomous vehicle is prone to instability when subjected to network attacks, constraints are set for the state variables, control variables, and control increments of the vehicle.

[0114] ​​First step, set the front wheel steering angle constraint and the increment constraint of each control step.

[0115] (23)

[0116] where, is the front wheel steering angle at time k, is the increment of the front wheel steering angle at time k, is the minimum value of the front wheel steering angle, is the maximum value of the front wheel steering angle, is the minimum value of the front wheel steering angle increment, is the maximum value of the front wheel steering angle increment.

[0117] Second step, set the constraint condition of the center of mass side slip angle on good road surface .

[0118] (24)

[0119] Third step, set the adhesion condition constraint, comprehensively consider the lateral and longitudinal acceleration of the vehicle.

[0120] (25)

[0121] where, is the longitudinal acceleration of the vehicle, is the lateral acceleration of the vehicle, is the adhesion coefficient, and g is the gravitational acceleration.

[0122] Step 2.5, set the target function for the reconstructed MPC controller and solve it.

[0123] First step, set the target function J up that meets the lateral tracking and stability as follows:

[0124] (26)

[0125] where, and are the weight matrices of the system prediction output deviation and the control increment, respectively, is the system prediction output, is the reference trajectory, and the reference trajectories of the yaw angle, yaw rate and lateral position, respectively.

[0126] Second step, solve the control increment sequence through the target function. After solving in each control period, a series of control input increments in the control time domain can be obtained:

[0127] (27)

[0128] wherein, is based on the control increment at the moment within the control step.

[0129] Thirdly, the first element in the sequence is taken as the actual control increment acting on the system, i.e.

[0130] (28)

[0131] wherein, is the control amount at the moment, is the control amount at the moment.

[0132] After entering the next control cycle, the above steps are repeated, and the tracking control of the desired trajectory is achieved through the cycle.

[0133] Step 3: Design the lower layer torque optimization distribution controller.

[0134] Step 3.1, the lower layer controller realizes the demand additional yaw moment and the longitudinal driving force required for speed following by distributing the torques of the four wheels. The present application designs the objective function with the minimum utilization rate of the four-wheel tires as the optimization goal:

[0135] (29)

[0136] wherein, c i is the weight coefficient corresponding to each wheel, F xi , F yi , F zi represent the longitudinal force, lateral force and vertical force of each wheel.

[0137] Since the longitudinal force and lateral force of the tire are coupled under extreme working conditions, the lateral stability margin can be improved by reducing the longitudinal force of the tire, and therefore, the optimization goal can be simplified as:

[0138] (30)

[0139] Step 3.2, set the constraint condition. While meeting the additional yaw moment and longitudinal force of the upper layer, the maximum output torque of the hub motor and the road adhesion condition are considered, and the longitudinal force constraint condition is expressed as:

[0140] (31)

[0141] wherein, is the maximum output torque of the hub motor. ​​​For additional yaw moment. B f For front wheel base, B r For rear wheel base.

[0142] Step 3.3, convert to quadratic programming solution. The above optimization problem is expressed in the form of norm as follows:

[0143] (32)

[0144] In the formula, ; ;

[0145] ;

[0146] , wherein is used to adjust the weight of the wheel longitudinal force;

[0147] is used to determine the priority of the expected additional yaw moment;

[0148] is the weight coefficient, when the weight coefficient is larger, the constraint condition is preferentially met to ensure the tracking accuracy and stability requirements.

[0149] At the same time, considering the road adhesion condition and motor performance, the maximum driving force of each wheel is defined as:

[0150] (33)

[0151] In the formula, is the adhesion coefficient of each wheel.

[0152] Therefore, the boundary condition of the control variable can be expressed as:

[0153] (34)

[0154] In the formula, is the maximum value of the control variable, is the minimum value of the control variable, is the maximum value of the wheel longitudinal force.

[0155] Step 4: design the lateral cooperative control rule under network attack as shown in Figure 3 . According to whether it is attacked by network, the front wheel steering angle , the expected yaw rate , the actual yaw rate , the lateral position deviation e c to determine the required additional yaw moment.​

[0156] The above is the overall framework of the intelligent electric vehicle lateral safety control under network attack. The controller calculates the required steering angle and wheel torque of the vehicle under the condition of known reference trajectory and state estimation information.

[0157] Figure 4 is a simulation verification of the intelligent vehicle lateral control method based on sensor measurement reconstruction under network attack. Among them, (a) is the trajectory comparison; (b) is the yaw angle comparison; (c) is the yaw rate comparison; (d) is the lateral acceleration comparison; (e) is the front wheel steering angle comparison; (f) is the wheel torque. The simulation is based on CarSim / Simulink, and the double lane shift working condition is adopted, the longitudinal vehicle speed is 15m / s, the road adhesion coefficient is 0.8, and the simulation compares the trajectory tracking and lateral stability control effect under the composite attack scene (including false data injection attack, denial of service attack and random packet loss). Among them, the fault-tolerant control is the control effect of the data disturbed by the network attack, and ATS-MPC and MFT-MPC are the control effects of sensor measurement reconstruction using different state estimation values. By comparing the trajectory, yaw angle, yaw rate, lateral acceleration, front wheel steering angle and wheel torque and other key indicators under different control strategies, it can be clearly seen that the driving stability and trajectory tracking ability of the vehicle in the network attack are significantly improved. These results provide strong technical support for the safe operation of intelligent vehicles in the network environment.

[0158] The above is a further detailed description of the preferred technical solution of the present application, and cannot be regarded as the specific implementation of the invention. For ordinary skilled persons in the art to which the present application belongs, without departing from the concept of the present application, simple deductions and substitutions can also be made, which should be regarded as the protection scope of the present application.

Claims

1. A method for lateral control of intelligent vehicles based on sensor measurement reconstruction under network attacks, characterized in that... Includes the following steps: Step 1: Establish a three-degree-of-freedom dynamic model describing the lateral, longitudinal, and yaw dynamic characteristics of the intelligent electric vehicle. Based on the transformation relationship of the longitudinal and lateral forces acting on the tires in the x-axis and y-axis directions in the vehicle coordinate system, simplify to obtain the nonlinear state-space equation. Discretize the equation, introduce process noise and measurement noise in state estimation, and add the influence of network attack signals and data packet loss on the sensor measurement output signal. Step 2: Design an upper-level MPC controller based on sensor measurement reconstruction; when the vehicle sensor system is subjected to malicious network attacks, the control law is switched in real time using the attack detection results and attack signal estimation results to reduce the impact of network attacks on the performance of the lateral control system of autonomous vehicles. Step 3: Design a lower-level torque optimization distribution controller; establish a torque optimization distribution method to control the vehicle's yaw moment; Step 4: Construct lateral collaborative control rules, taking into account network attack factors, and utilize the vehicle's redundancy yaw stability to reduce lateral tracking error. By constructing lateral collaborative control rules, the upper-level MPC controller and the lower-level torque optimization distribution controller can work together to jointly address the impact of network attacks on the vehicle's lateral control.

2. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 1, characterized in that... In step 1, establishing a three-degree-of-freedom dynamic model describing the lateral, longitudinal, and yaw dynamic characteristics of the intelligent electric vehicle specifically includes the following steps: 1.1) Establish a three-degree-of-freedom dynamic model that accurately characterizes the behavior mechanism of intelligent electric vehicles. This model includes the lateral, longitudinal, and yaw dynamic characteristics of the vehicle. 1.2) Establish the conversion relationship between the longitudinal and lateral forces acting on the tire in the x-axis and y-axis directions in the vehicle coordinate system; 1.3) Simplify to obtain the nonlinear state-space equation of the intelligent electric vehicle; 1.4) The above nonlinear state-space equations are approximately discretized using the first-order difference quotient method to obtain the nonlinear discrete state equations; 1.5) Introduce process noise and measurement noise to reflect the uncertainties in the actual driving process, and establish the state equation and measurement equation for state estimation based on the nonlinear discrete state equation; 1.6) Introduce network attack signals and establish measurement equations for sensors subjected to network attacks; 1.7) Introduce random data loss and establish a measurement equation for data loss caused by the sensor.

3. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 2, characterized in that... The establishment of a three-degree-of-freedom dynamic model describing the lateral, longitudinal, and yaw dynamic characteristics of an intelligent electric vehicle specifically includes the following steps: Step 1.1: According to Newton's second law, obtain the force balance equations for longitudinal, lateral, and yaw forces, as well as the conversion equations between the vehicle coordinate system and the inertial coordinate system: In the formula, , These are the distances from the center of mass to the front and rear axles, respectively. For vehicle curb weight Let be the moment of inertia of the vehicle about the z-axis. Let be the lateral speed of the vehicle. Let be the longitudinal speed of the vehicle. For the vehicle's lateral acceleration, For the longitudinal acceleration of the vehicle, The yaw angle of the vehicle. Let yaw rate be the vehicle's angular velocity. The lateral velocity is in the geodetic coordinate system. This represents the longitudinal velocity in the geodetic coordinate system. The force is in the x-axis direction of the front wheel. The force in the x-axis direction of the rear wheel. The force is in the y-axis direction of the front wheel. This refers to the force along the y-axis of the rear wheel; Step 1.2, the conversion relationship between the resultant force and the longitudinal and lateral forces acting on the tire in the x-axis and y-axis directions in the vehicle coordinate system is as follows: In the formula, The force is in the x-axis direction of the front wheel. The force in the x-axis direction of the rear wheel. The force is in the y-axis direction of the front wheel. The force in the y-axis direction of the rear wheel. For the longitudinal force of the front wheel, For the longitudinal force of the rear wheel, The lateral force is from the front wheel. The lateral force is from the rear wheel. The steering angle of the front wheels; Assuming the tire lateral slip characteristics of an intelligent electric vehicle are within a linear range, the longitudinal force and lateral force of the tire can be obtained using the following formulas: In the formula, For the longitudinal force of the front wheel, For the longitudinal force of the rear wheel, The lateral force is from the front wheel. This refers to the lateral force on the rear wheel; For the longitudinal stiffness of the front tire; For the longitudinal stiffness of the rear tire; This refers to the lateral stiffness of the front tire; For the lateral stiffness of the rear tire; This refers to the front tire slip ratio; Rear tire slip ratio; This refers to the front tire slip angle; This refers to the rear tire slip angle; By simplifying the calculation, the formula for calculating the tire slip angle is obtained: In the formula, This refers to the front tire slip angle. The rear tire slip angle, Let be the lateral speed of the vehicle. Let yaw rate be the vehicle's angular velocity. The longitudinal speed of the vehicle; , These are the distances from the center of mass to the front and rear axles, respectively. This refers to the front wheel deflection angle; Step 1.3 yields the following set of differential equations for the nonlinear intelligent electric vehicle dynamics model: In the formula, For vehicle curb weight This refers to the vehicle's longitudinal acceleration. The lateral acceleration of the vehicle. Let yaw acceleration be the acceleration of the vehicle. Let be the lateral speed of the vehicle. Let yaw rate be the vehicle's angular velocity. For the longitudinal stiffness of the front tire; For the longitudinal stiffness of the rear tire, For the lateral stiffness of the front tire, For the lateral stiffness of the rear tire, The front tire slip ratio, Rear tire slip ratio, Let be the moment of inertia of the vehicle about the z-axis. Let the longitudinal velocity of the vehicle be in the geodetic coordinate system. Let be the lateral velocity of the vehicle in the geodetic coordinate system; Let the state variable be The control quantity is The nonlinear dynamic model can then be abstracted into a nonlinear state-space equation, as shown in the following equation: in, Represents the state variable at time t. Represents the control variable at time t. This represents the output variable at time t. This represents the first differential of the state variable at time t. Represents the abstract obtained about and The function, express and The function; Step 1.4: Approximate discretization of the above nonlinear state-space equations using the first-order difference quotient method yields the nonlinear discrete state equations: In the formula, for State quantity at any given time. for State quantity at any given time. Let k be the control variable at time k. The sampling period; Step 1.5, considering process noise and measurement noise, the above nonlinear discrete state equations and measurement equations for the autonomous vehicle are integrated into the following form: in, , Indicates process noise. Indicates measurement noise; Step 1.6: Establish sensor measurement equations that accurately characterize cyberattacks on autonomous vehicles; sensor data at time k under cyberattack. The equation is described as follows: In the formula, It is an output variable. It is malicious data injected by the attacker. A sensor attack matrix set up for attackers; Establish the nonlinear discrete state equations and measurement equations for the vehicle under network attacks: In the formula, Indicates process noise. Indicates measurement noise. It is an output variable. It is malicious data injected by the attacker. A sensor attack matrix set up for attackers; express State quantity at any given time. express State quantity at any given moment; Step 1.7, considering random packet loss, the nonlinear discrete state equation and measurement equation for the autonomous vehicle are as follows: In the formula, , , For sensor measurement output, These are independent random variables introduced to describe the random measurement delays and random packet loss phenomena of the system; when the variable This indicates that the vehicle communication network successfully transmitted data without packet loss, and the system is in a closed-loop state; when the variable When this occurs, it indicates that data transmission failed through the vehicle communication network, resulting in packet loss, and the system is in an open-loop state; the state transition probability matrix of the two-state Markov chain is represented as: ， ， In the formula, p ij This represents the state transition probability of a Markov chain. for Independent random variables at time t, for Independent random variables at time t.

4. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 1, characterized in that... In step 2, the design is based on an upper-layer MPC controller for sensor measurement reconstruction, specifically including: 2.1) Assume that the vehicle state information obtained through the vehicle sensor system has been affected by network attacks and packet loss, and at the same time, obtain an accurate state estimate through the state estimation method; 2.2) Match the lateral control system model affected by the cyberattack with the nominal lateral control system model; 2.3) Introduce state estimates affected by network attacks and packet loss to reconstruct the output state prediction model in the upper-layer MPC controller; 2.4) Set constraints on the vehicle's state variables, control variables, and control increments; 2.5) Set the objective function for the reconstructed upper-level MPC controller and solve it to obtain the optimal control strategy.

5. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 4, characterized in that... In step 2.4), setting constraints on the vehicle's state variables, control variables, and control increments includes: The first step is to set the front wheel steering angle constraint and the incremental constraints for each control step; In the formula, This represents the front wheel steering angle at time k. This represents the increment of the front wheel steering angle at time k. This represents the minimum front wheel steering angle. This indicates the maximum value of the front wheel steering angle. This represents the minimum value of the front wheel steering angle increment. This represents the maximum value of the front wheel steering angle increment; The second step is to set the center of gravity sideslip angle on a good road surface. Constraints; The third step is to set attachment condition constraints, taking into account the vehicle's lateral and longitudinal accelerations. In the formula, For the longitudinal acceleration of the vehicle, For the vehicle's lateral acceleration, denoted as the adhesion coefficient, and g as the acceleration due to gravity.

6. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 4, characterized in that... In step 2.5), setting and solving the objective function for the reconstructed upper-layer MPC controller specifically includes: The first step is to set the objective function J that satisfies lateral tracking and stability. up for: In the formula, and These are the weight matrices for the system's predicted output deviation and control increment, respectively. To predict the output of the system, The reference trajectories are the yaw angle, yaw angular velocity, and lateral position reference trajectories, respectively. The second step is to solve for the control increment sequence using the objective function; after solving within each control cycle, a series of control input increments in the control time domain are obtained. : In the formula, For based on The control increment is always within the control step size; The third step is to extract the first element from the sequence. As an actual control increment acting on the system, that is In the formula, for The amount of control at any given moment for The amount of control at any given moment; After entering the next control cycle, the above steps are repeated, and so on, to achieve tracking control of the desired trajectory.

7. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 1, characterized in that... In step 3, the design of the lower-level torque optimization distribution controller specifically includes: 3.1) Design an objective function with the goal of minimizing the overall adhesion utilization rate of the four wheels; 3.2) Set constraints based on road surface adhesion conditions and hub motor performance; 3.3) The optimization problem is transformed into a quadratic programming problem to solve for the torque of the four wheels.

8. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 7, characterized in that... In step 3.1), the objective function designed with minimizing the overall adhesion utilization rate of the four wheels as the optimization objective is as follows: The lower-level controller distributes the torque of the four wheels to achieve the additional yaw moment required by the upper-level solution. and the longitudinal driving force required for vehicle speed following The objective function is designed with minimizing the overall adhesion utilization rate of the four tires as the optimization objective: In the formula, c i F represents the weighting coefficient for each wheel. xi F yi F zi This represents the longitudinal, lateral, and vertical forces on each wheel; Since the longitudinal and lateral forces of the tire are coupled under extreme conditions, reducing the longitudinal force of the tire increases the lateral stability margin. Therefore, the optimization objective can be simplified to: In the formula, c i F represents the weighting coefficient for each wheel. xi F yi F zi This indicates the longitudinal, lateral, and vertical forces on each wheel.

9. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 7, characterized in that... In step 3.2), the constraint conditions are set according to the road surface adhesion conditions and the performance of the hub motor: while satisfying the additional yaw moment and longitudinal force of the upper layer, the maximum output torque of the hub motor and the road surface adhesion conditions are considered. The longitudinal force constraint condition is expressed as follows: In the formula, This represents the maximum output torque of the hub motor; To add yaw moment; B f B is the front wheelbase. r This refers to the rear wheelbase.

10. The intelligent vehicle lateral control method based on sensor measurement reconstruction under network attacks as described in claim 1, characterized in that... In step 4, the construction of lateral collaborative control rules is based on whether there is a network attack and the front wheel steering angle. Desired yaw rate Actual yaw rate Lateral position deviation e c Determine the required additional yaw moment.

Images