An evolutionary algorithm-assisted secure PID controller design method considering network hybrid attacks

Abstract

The present invention relates to an evolutionary algorithm-assisted safety PID controller design method considering network hybrid attacks, which solves the problems of slow response speed, insufficient robustness and heavy computational burden of traditional methods. The specific steps include: establishing a discrete time network control system model, defining state vector equations and output equations; defining a hybrid attack network model; designing an observer-based PID controller, and using GA genetic algorithm and LMI linear matrix inequality to obtain the PID controller gain and observer gain to be designed; using an event trigger mechanism to determine whether to transmit the output of the PID controller to an actuator; and realizing stabilization of the UAV system according to the output of the actuator. The present invention improves the robustness and response speed of the UAV control system when facing hybrid attacks, reduces the computational and communication burdens, and ensures the safe and stable operation of the UAV in a complex environment.

Description

Technical Field

[0001] The present invention relates to the technical field of network control systems, and in particular to a design method for a secure PID controller assisted by an evolutionary algorithm taking network hybrid attacks into consideration. Background Art

[0002] With the widespread application of network control systems (NCS) in industrial automation, smart grids, and drone control, system security issues have become particularly important. Hybrid network attacks, such as false data injection (FDI) and replay attacks, seriously threaten the stability and performance of NCS. Traditional control methods cannot fully cope with these challenges, so a new controller design method is needed to improve the robustness and security of the system.

[0003] PID controllers are widely used in industrial control due to their simplicity and practicality. However, traditional PID controllers show obvious limitations when facing complex network attacks. In recent years, various evolutionary algorithms (such as genetic algorithms (GAs)) have been widely used in optimizing controller design. Combining evolutionary algorithms with linear matrix inequalities (LMIs) is expected to improve the robustness and performance of controllers.

[0004] As an emerging network control system, unmanned aerial vehicles (UAVs) have been widely used in military, logistics, agriculture, and environmental monitoring. The complexity of UAV control systems and the diversity of their tasks make them one of the main targets of cyber attacks. Cyber ​​attacks not only affect the flight stability of UAVs, but may also lead to mission failures and even serious safety accidents. Therefore, designing a UAV controller that can effectively respond to hybrid network attacks is of great significance to improving the safety and reliability of UAVs.

[0005] Traditional control methods for cyber attacks on drone control systems have the following limitations:

[0006] Slow response speed: The traditional PID control method has insufficient response speed when facing rapidly changing external interference and attacks, and it is difficult to ensure the real-time performance of the system;

[0007] Insufficient robustness: Traditional methods do not fully consider the uncertainty and nonlinearity caused by attacks during the design process, which makes the system prone to instability when attacked;

[0008] Heavy computational burden: UAV control systems often need to process a large amount of sensor data. Traditional control methods increase the computational and communication burden while ensuring performance.

[0009] In traditional UAV control problems, controller design often involves nonlinear terms or non-convex constraints. In order to make these problems solvable, it is usually necessary to adopt a conservative method to transform these nonlinear or non-convex terms into linear terms, and then use the linear matrix inequality (LMI) method to solve the controller parameters. In recent years, various evolutionary algorithms have made significant progress in solving the optimization of non-convex problems. Combining evolutionary algorithms with LMI methods for controller design can effectively reduce conservatism and improve system performance. Summary of the invention

[0010] In view of the shortcomings of the prior art, the present invention proposes an evolutionary algorithm-assisted safety PID controller design method considering network hybrid attacks. By combining the GA genetic algorithm and the LMI linear matrix inequality, the method aims to improve the robustness and response speed of the UAV control system in the face of network hybrid attacks, reduce the calculation and communication burden, and ensure the safe and stable operation of the UAV in a complex environment.

[0011] In order to solve the above technical problems, the present invention provides the following technical solutions, a method for designing an evolutionary algorithm-assisted secure PID controller considering network hybrid attacks, which specifically includes the following steps:

[0012] S1. Establish a discrete time network control system model and define the state vector equation and output equation. The control system model includes: UAV system, sensor, hybrid attack network model, observer, PID controller, event trigger mechanism and actuator;

[0013] S2. Define a hybrid attack network model; the drone system state vector is input into the hybrid attack network model through a sensor, and after hybrid attack simulation, it is input into the observer; the hybrid attack includes false data injection attack and replay attack, and the probability of attack occurrence is described by Bernoulli distribution;

[0014] S3. Design an observer-based PID controller, including three parts: proportional P, integral I, and differential D. Use GA genetic algorithm and LMI linear matrix inequality to obtain the PID controller gain and observer gain to be designed. The PID controller is based on the state estimation vector of the UAV system given by the observer. is the input;

[0015] S4, using the event trigger mechanism to determine whether the output of the PID controller Transmitted to the actuator;

[0016] S5. When the event trigger mechanism is activated, the output of the PID controller is transmitted to the actuator, and the UAV system is stabilized through the actuator.

[0017] Furthermore, the discrete-time network control system model established in step S1 is specifically:

[0018] ;

[0019] in, express k The state vector of the UAV system at the moment, express k+ The state vector of the UAV system at time 1, The control signal indicating the latest time, express k The external interference to the drone system at any time, express k The output of the sensor at this moment, express k The control output of the UAV system at all times; It is a parameter matrix obtained by discretization, which describes the transformation of the system's dynamic equations from continuous time form to the state vector in discrete time form.

[0020] Furthermore, the formula of the hybrid attack network model in step S2 is expressed as:

[0021] ;

[0022] in, represents the total output of the hybrid attack network model, express k The output of the sensor at the time; using Bernoulli variables Indicates whether a mixed attack is received. Indicates that it was attacked by a combination of When it is not attacked by mixed attacks, it satisfies ; represents the probability of an event occurring, is a constant value in [0,1];

[0023] Represents a mixed attack signal, and the formula is:

[0024] ;

[0025] in, is a false data injection attack signal that satisfies the constraints:

[0026] ;

[0027] in, is the UAV system state estimation vector given by the observer, is a real matrix of known dimension, is a given constant; Indicates replay attack signal, In a specific range: ; is another Bernoulli variable that also satisfies , represents the probability of an event occurring, is a constant value in [0,1].

[0028] Furthermore, step S3 specifically includes the following steps:

[0029] S31. Convert the PID controller design problem into an LMI linear matrix inequality problem and use the CCL cone complement linearization algorithm to obtain a set of feasible PID controller gains. and the feasible solution of the observer gain , and then the population initialization range of the GA genetic algorithm can be obtained;

[0030] S32, combine GA genetic algorithm and LMI linear matrix inequality to obtain the observer gain to be designed L and the PID controller gains, which are the corresponding gains of the three parts of the controller ;

[0031] S33, the UAV system state estimation vector given by the observer As input, the output of the PID controller is .

[0032] Furthermore, step S31 specifically includes the following steps:

[0033] S311. Apply Schur's complement lemma to the matrix inequality with nonlinear terms of the PID controller, and the corresponding matrix inequality is as follows:

[0034] ;

[0035] in,

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] in, is a given constant, is a given constant, , is the positive definite matrix to be found, is the identity matrix, , is a given constant greater than 0, is the performance indicator to be sought The value of

[0041] ;

[0042] in, is the parameter matrix, is a real matrix of known dimension;

[0043] S312: Apply the CCL cone complement linearization algorithm to the linear matrix inequality obtained in S311 and other linear matrix inequalities to obtain a set of feasible PID controller gains. and the feasible solution of the observer gain , and then obtain the population initialization range of the GA genetic algorithm.

[0044] Further, step S32 specifically includes the following steps:

[0045] S321, GA genetic algorithm population initialization when each individual initialization range is , ,

[0046] in, Indicates the proportional coefficient corresponding to each gain, which is given by experience;

[0047] S322, randomly generate an initial population from the initialization range obtained in S321;

[0048] S323, using the LMI tool to solve the linear matrix inequality group for each individual, if there is a solution, calculate the corresponding fitness function value; if there is no solution, assign a value to the fitness value of the individual;

[0049] S324, executing selection, crossover, mutation, and elite retention GA operations;

[0050] S325. If the maximum number of iterations is reached, the optimal individual gain obtained during the iteration process is saved. and observer gain , otherwise return to step S323.

[0051] Further, in step S33, the UAV system state estimation vector given by the observer is and the output of the PID controller The formula is:

[0052] ;

[0053] in, represents the total output of the hybrid attack network model, express k The UAV system state estimation vector at time t is: Respectively k+ 1 moment and k- The estimated state vector of the UAV system at time 1, Indicates from k Time has come k- The accumulation of the UAV system state estimation vector at the moment 1 is given by the observer. Indicates the control signal at the latest time, parameter is a given positive scalar; Represents the output of the PID controller; Indicates the PID controller gain to be designed, corresponding to the proportion P ,integral I ,differential D Three parts, L represents the observer gain to be designed.

[0054] Furthermore, the event triggering mechanism adopted in step S4 is specifically:

[0055] Design event trigger function , its formula is expressed as:

[0056] ;

[0057] in, is a known scalar, , Indicates the output of the PID controller at the latest moment; when the trigger function When the event trigger mechanism is activated, the output of the PID controller at the latest moment is As the input of the actuator, get the latest control signal , the formula is:

[0058] ;

[0059] When the event trigger mechanism is not in effect, the latest control signal remains unchanged from the control signal at the previous moment, that is, .

[0060] By means of the above technical solution, the present invention provides an evolutionary algorithm-assisted secure PID controller design method considering network hybrid attacks, which has at least the following beneficial effects:

[0061] (1) The present invention optimizes the design of the PID controller through a genetic algorithm, and the system can automatically adjust the controller parameters to adapt to different working conditions and external interference. This optimization not only improves the system's resistance to network hybrid attacks (such as false data injection and replay attacks), but also enhances the system's stability and security when facing attacks. The optimized PID controller can quickly respond to changes in system status and quickly adjust to ensure system stability when an attack occurs.

[0062] (2) The present invention introduces an event trigger mechanism to effectively reduce the transmission frequency of control signals. Compared with conventional periodic communication, the event trigger mechanism transmits data only when the system state changes significantly, thereby reducing the communication burden and enabling the system to respond to external changes and attacks more quickly.

[0063] (3) The present invention combines evolutionary algorithms and linear matrix inequality (LMI) methods to design controllers and optimize controller parameters. Evolutionary algorithms can find global optimal solutions in non-convex optimization problems, while LMI methods ensure the stability and performance of controller design. This combination effectively reduces the computational and communication burden of the system and improves the efficiency and reliability of the system. The introduction of evolutionary algorithms and LMI methods not only improves the flexibility and accuracy of controller design, but also reduces the performance loss caused by conservatism in traditional methods. The optimized controller design is more effective and reliable, and can provide stable performance in various complex control environments.

[0064] (4) The present invention is applicable to complex UAV control systems and can improve the safety and reliability of the system in a variety of mission environments. UAV systems are usually faced with various complex environments and attack threats. The optimized controller design can effectively cope with these challenges and ensure flight safety and stability of mission completion. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0066] Figure 1 is a general flow chart of an embodiment of the present invention;

[0067] Figure 2 It is a fitness value iteration curve diagram of the genetic algorithm of an embodiment of the present invention;

[0068] Figure 3 A schematic diagram of a release interval of an event triggering mechanism according to an embodiment of the present invention;

[0069] Figure 4 It is an open-loop state trajectory curve diagram of the UAV system according to an embodiment of the present invention;

[0070] Figure 5 It is an open-loop state trajectory curve diagram of the UAV system according to an embodiment of the present invention;

[0071] Figure 6 Flow chart of the genetic algorithm of the embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to more clearly show the purpose, features and advantages of the present invention, further detailed description is given in conjunction with the accompanying drawings and specific implementation methods. The present application uses technical means to solve technical problems, and the process of achieving technical effects can be completed by program instructions related hardware. Therefore, the present invention can be implemented in full hardware, full software, or a combination of software and hardware.

[0073] Those skilled in the art will appreciate that all or part of the steps in the implementation method of the present invention may be completed by instructing the relevant hardware through a computer-executable program. The embodiments of the present application may include a computer program product implemented on a computer-usable storage medium, such as a disk storage, CD-ROM, optical storage, etc., which contains a computer-executable program code.

[0074] This way of expression can not only reduce the repetitiveness of invention patent applications, but also better describe the flexibility and application scope of technical implementation.

[0075] Reference Figure 1-Figure 6 , a specific implementation method of this embodiment is given. This embodiment uses a genetic algorithm GA combined with a linear matrix inequality LMI to design a PID controller. The genetic algorithm GA can find the global optimal solution in non-convex optimization problems, while the linear matrix inequality LMI ensures the stability and performance of the controller design. The combination of the two optimizes the parameters of the controller, so that it can quickly respond to changes in the state of the drone system and quickly adjust to ensure the stability of the system when a hybrid attack occurs; in addition, the introduced event trigger mechanism reduces the communication burden and enables the system to respond to external changes and attacks more quickly; therefore, the present invention is suitable for complex drone control systems and can improve the safety and reliability of the system in a variety of mission environments.

[0076] Please refer to Figure 2 This embodiment proposes an evolutionary algorithm-assisted secure PID controller design method considering network hybrid attacks, and the method includes the following steps:

[0077] S1. Establish a discrete time network control system model and define the state vector equation and output equation. The control system model includes: UAV system, sensor, hybrid attack network model, observer, PID controller, event trigger mechanism and actuator;

[0078] As a preferred implementation of step S1, the specific process is:

[0079] The discrete-time network control system model established is as follows:

[0080] ;

[0081] in, express k The state vector of the UAV system at the moment, express k+1 The state vector of the UAV system at the moment, The control signal indicating the latest time, express k The external interference to the drone system at any time, express k The output of the sensor at this moment, express k The control output of the UAV system at all times; is a parameter matrix obtained through discretization, which describes the system's state vector converted from the dynamic equation in continuous time to the discrete time form; the specific formula for discretization is: ;

[0082] in, They are the speed, track angle, heading angle, normal overload, roll angle and thrust of the drone. They are gravity acceleration, maximum power, a coefficient related to aerodynamics, mass of the drone, dynamic pressure, wing area, zero lift drag coefficient and normal overload. The discrete time system sampling interval is selected as 0.02s, and the c2d function is used in MATLAB. The discretization method is the zero-order hold method. The corresponding discrete time state space model can be obtained from the discretized discrete time state space model. Equal parameter matrix.

[0083] In this embodiment, the present invention can obtain the desired discrete system parameters by discretizing the continuous system parameters, so as to verify the practical application effect of the method of the present invention.

[0084] S2. Define a hybrid attack network model; the drone system state vector is input into the hybrid attack network model through a sensor, and after hybrid attack simulation, it is input into the observer; the hybrid attack includes false data injection attack and replay attack, and the probability of attack occurrence is described by Bernoulli distribution;

[0085] As a preferred implementation of step S2, the hybrid attack network model is specifically expressed as:

[0086] ;

[0087] in, represents the total output of the hybrid attack network model, express k The output of the sensor at the time; using Bernoulli variables Indicates whether a mixed attack is received. Indicates that it was attacked by a combination of When it is not attacked by mixed attacks, it satisfies ; represents the probability of an event occurring, is a constant value in [0,1];

[0088] Represents a mixed attack signal, and the formula is:

[0089] ;

[0090] in, is a false data injection attack signal that satisfies the constraints:

[0091] ;

[0092] in, is the UAV system state estimation vector given by the observer, is a real matrix of known dimension, is a given constant; Indicates replay attack signal, In a specific range: ; is another Bernoulli variable that also satisfies , represents the probability of an event occurring, is a constant value in [0,1].

[0093] In this embodiment, the present invention simulates false data injection attacks and replay attacks in the drone control system to analyze their impact on system stability.

[0094] S3. Design an observer-based PID controller, including three parts: proportional P, integral I, and differential D. Use GA genetic algorithm and LMI linear matrix inequality to obtain the PID controller gain and observer gain to be designed. The PID controller is based on the state estimation vector of the UAV system given by the observer. is the input;

[0095] As a preferred implementation of step S3, the following steps are specifically included:

[0096] S31. Convert the PID controller design problem into an LMI linear matrix inequality problem and use the CCL cone complement linearization algorithm to obtain a set of feasible PID controller gains. and the feasible solution of the observer gain , and then the population initialization range of the GA genetic algorithm can be obtained;

[0097] More specifically, step S31 includes:

[0098] S311. Apply Schur's complement lemma to the matrix inequality with nonlinear terms of the PID controller, and obtain the corresponding matrix inequality as follows:

[0099] ;

[0100] in,

[0101] ;

[0102] ;

[0103] ;

[0104] ;

[0105] in, is a given constant, is a given constant, , is the positive definite matrix to be found, is the identity matrix, , is a given constant greater than 0, is the performance indicator to be sought The value of

[0106] ;

[0107] in, is the parameter matrix, is a real matrix of known dimension;

[0108] S312: Apply the CCL cone complement linearization algorithm to the linear matrix inequality obtained in S311 and other linear matrix inequalities to obtain a set of feasible PID controller gains. and the feasible solution of the observer gain , and then obtain the population initialization range of the GA genetic algorithm; in this embodiment, the initial value obtained by using the CCL algorithm can provide a good population initialization range for our subsequent use of the GA algorithm, making our search space more effective;

[0109] S32, combine GA genetic algorithm and LMI linear matrix inequality to obtain the observer gain to be designed L and the PID controller gains, which are the corresponding gains of the three parts of the controller ;

[0110] More specifically, step S32 includes:

[0111] S321, GA genetic algorithm population initialization when each individual initialization range is , ,

[0112] in, Indicates the proportional coefficient corresponding to each gain, which is given by experience;

[0113] S322, randomly generate an initial population from the initialization range obtained in S321;

[0114] S323, use the LMI tool to solve the linear matrix inequality group for each individual, if there is a solution, calculate the corresponding fitness function value; if there is no solution, assign a value to the fitness value of the individual; assigning a value here is to facilitate drawing such as Figure 2 The fitness curve graph shown usually takes a value larger than other calculated fitness function values;

[0115] S324, executing selection, crossover, mutation, and elite retention GA operations;

[0116] S325. If the maximum number of iterations is reached, the optimal individual gain obtained during the iteration process is saved. and observer gain , otherwise return to step S323;

[0117] In this embodiment, the LMI method is combined with the GA genetic algorithm to reduce the conservatism in the controller design and effectively improve the performance of the UAV system; the introduction of the genetic algorithm can well solve the non-convex and nonlinear terms in the sufficient conditions of the PID controller, and at the same time, compared with the traditional PID control, it can reduce the value of the infinite performance index H, thereby optimizing the performance of the controller design and improving the robustness and adaptability of the system;

[0118] S33, the UAV system state estimation vector given by the observer As input, the output of the PID controller is ;

[0119] More specifically, in step S33, the UAV system state estimation vector given by the observer is and the output of the PID controller The formula is:

[0120] ;

[0121] in, represents the total output of the hybrid attack network model, express k The UAV system state estimation vector at time t is: Respectively k+ 1 moment and k- The estimated state vector of the UAV system at time 1, Indicates from k Time has come k- The accumulation of the UAV system state estimation vector at the moment 1 is given by the observer. Indicates the control signal at the latest time, parameter is a given positive scalar; Represents the output of the PID controller; Indicates the PID controller gain to be designed, corresponding to the proportion P ,integral I ,differential D Three parts, L represents the observer gain to be designed;

[0122] In step S3 of this embodiment, the present invention uses an observer-based PID controller for the UAV system, and combines genetic algorithm and LMI method to optimize controller parameters, thereby providing the UAV system with higher control accuracy, stability and adaptability, while reducing the cost and complexity of the system.

[0123] S4, using the event trigger mechanism to determine whether the output of the PID controller Transmitted to the actuator;

[0124] As a preferred implementation of step S4, the event triggering mechanism is expressed as:

[0125] ;

[0126] in, is a known scalar, take , , Indicates the output of the PID controller at the latest moment; when the trigger function When the event trigger mechanism is activated, the output of the PID controller at the latest moment is As the input of the actuator, get the latest control signal , the formula is:

[0127] ;

[0128] When the event trigger mechanism is not in effect, the latest control signal remains unchanged from the control signal at the previous moment, that is, ;

[0129] In this embodiment, drones usually rely on batteries for power supply, and energy consumption is an important consideration in the design; then the event trigger mechanism reduces the energy consumption required for frequent communication and calculation, prolongs the flight time and endurance, and only performs necessary operations when the state changes, effectively reducing energy consumption. In addition, the event trigger mechanism can also improve the response speed and system stability to adapt to complex and changing environmental conditions.

[0130] S5. After the event trigger mechanism determines, the output of the PID controller Transmitted to the actuator, the actuator is used to stabilize the drone system.

[0131] Combination Figure 2-Figure 6 , giving the actual verification effect of this embodiment on the method proposed in the present invention.

[0132] First, the performance index obtained by defining fitness is described as The value of fitness, i.e., performance index Value The lower it is, the better the performance of the PID controller is. The fitness curve obtained by multiple iterations is shown in Figure 2 As shown in the figure, the dark green straight line represents the performance index directly solved by the traditional method through the CCL cone complement linearization algorithm. Minimum value The red solid line represents the performance index of the best individual in the iterative optimization process of the GA genetic algorithm proposed in this invention. Value It can be seen that after more than ten iterations, the fitness of the method proposed by the present invention is significantly lower than that of the traditional method, that is, it has better performance.

[0133] The optimized PID controller is applied to the UAV system, where the event trigger mechanism releases the interval as Figure 4 As shown, it can be seen that the event-triggered communication mechanism we designed can reduce the transmission frequency of our control signals and effectively reduce the burden of UAV communication.

[0134] Figure 5 The state trajectory of the UAV system in the open loop is shown, that is, the change of the UAV system state over time without the output of the PID controller. The three curves in the figure represent different state variables. Over time k Evolution: (Blue curve): represents the change of the drone's speed. It can be seen that in the open loop, the speed V continues to decrease and eventually approaches a negative value, which indicates that the uncontrolled system will lose speed and show an unstable state; (Black curve): represents the track angle In the open loop, the track angle changes slowly and oscillates within a certain range, eventually stabilizing, but there is no obvious control effect; (Yellow curve): represents the track azimuth In the open-loop case, the track azimuth changes gradually, but the change is small, indicating that without control, the change of the track azimuth is not significant; overall, in the open-loop case, the speed of the system tends to be out of control, and although the track angle and track azimuth do not change significantly, they still show a trend of lack of control.

[0135] As a comparison, Figure 6 The state trajectory of the UAV system under closed-loop control is shown, that is, the change of the state of the UAV system over time under the action of the PID controller. The three curves in the figure also represent different state variables. Over time k Evolution: (Blue curve): Under closed-loop control, the speed v quickly rises to a peak value, then decreases and approaches zero, which indicates that the PID controller can quickly correct the speed deviation and restore the system to a stable state; (Black curve): Track angle Under closed-loop control, it quickly reaches a small positive peak value, then decreases rapidly and approaches zero, indicating that the PID controller effectively corrects the deviation of the track angle and the system returns to stability; (Yellow curve): Track azimuth Under closed-loop control, the change is very small and basically remains near zero, indicating that the PID controller effectively maintains the stability of the track azimuth. Overall, under the closed-loop condition, through the action of the PID controller, the system speed and track angle are quickly corrected and stabilized, and the track azimuth is maintained near the preset value, which significantly improves the stability and response speed of the system.

[0136] In summary, the evolutionary algorithm-assisted safety PID controller design method considering network hybrid attacks proposed in this invention optimizes the parameter configuration of the PID controller by combining the genetic algorithm GA and the linear matrix inequality LMI, making it applicable to complex UAV control systems, and can respond quickly and accurately to hybrid attacks, thereby improving the safety and reliability of the control system in a variety of mission environments. UAV systems are usually faced with various complex environments and attack threats, and the optimized PID controller can effectively respond to these challenges and ensure flight safety and the stability of mission completion.

[0137] In this specification, the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" refer to at least one embodiment or example described in combination with specific features, structures, materials or characteristics. These described specific features, structures, materials or characteristics can be combined in an appropriate manner in one or more embodiments or examples. In addition, the technician can combine and combine different embodiments or examples and their features described in this specification without contradiction.

[0138] The logic and / or steps shown in the flowchart or described in other ways can be regarded as a sequence of executable instructions for implementing logical functions. These instructions can be implemented in any computer-readable medium for use by instruction execution systems, devices or equipment. These systems, devices or equipment include processor systems or other systems capable of receiving and executing instructions.

[0139] The above embodiments have described the principles and implementations of the present invention in detail, and have used specific examples to illustrate the working principles thereof. These examples are only used to help understand the method of the present invention and its core concept. Meanwhile, according to the concept of the present invention, the actual implementation methods and application scope may vary. Therefore, the contents of this specification should not be construed as limiting the present invention.

Claims

1. An evolutionary algorithm-assisted secure PID controller design method considering network hybrid attacks, characterized in that: The specific steps include: S1. Establish a discrete time network control system model and define the state vector equation and output equation. The control system model includes: UAV system, sensor, hybrid attack network model, observer, PID controller, event trigger mechanism and actuator; S2. Define a hybrid attack network model; the drone system state vector is input into the hybrid attack network model through a sensor, and after hybrid attack simulation, it is input into the observer; the hybrid attack includes false data injection attack and replay attack, and the probability of attack occurrence is described by Bernoulli distribution; The formula of the hybrid attack network model is expressed as: Among them, y k represents the total output of the hybrid attack network model, Represents the output of the sensor at time k; using the Bernoulli variable β 1k Indicates whether a mixed attack is received, β 1k =1 means it is attacked by mixed attacks, β 1k =0 means no mixed attack, satisfying Prob() represents the probability of an event occurring. is a constant value in [0,1]; η k Represents a mixed attack signal, and the formula is: in, is a false data injection attack signal that satisfies the constraints: in, is the UAV system state estimation vector given by the observer, T is a real matrix of known dimension, and ρ is a given constant; represents the replay attack signal, τ k In a specific range: β 2k is another Bernoulli variable that also satisfies Prob() represents the probability of an event occurring. is a constant value in [0,1]; S3. Design an observer-based PID controller, including three parts: proportional P, integral I, and differential D. Use GA genetic algorithm and LMI linear matrix inequality to obtain the PID controller gain and observer gain to be designed. The PID controller is based on the state estimation vector of the UAV system given by the observer. is the input; S4, using the event trigger mechanism to determine whether to output the PID controller k Transmitted to the actuator; S5, when the event trigger mechanism is activated, the output of the PID controller k Transmitted to the actuator, the actuator is used to stabilize the drone system.

2. According to claim 1, the method for designing an evolutionary algorithm-assisted secure PID controller considering network hybrid attacks is characterized in that: The discrete-time network control system model established in step S1 is specifically: Among them, x k represents the state vector of the UAV system at time k, x k+1 represents the state vector of the UAV system at time k+1, u k Represents the control signal at the latest moment, ω k represents the external interference to the UAV system at time k, represents the output of the sensor at time k, z k represents the control output of the UAV system at time k; A, B, C, D, and E are parameter matrices obtained through discretization, which describe the state vector of the system converted from the dynamic equation in continuous time form to the discrete time form.

3. The method for designing an evolutionary algorithm-assisted secure PID controller considering network hybrid attacks according to claim 1 is characterized in that: Step S3 specifically includes the following steps: S31. Convert the PID controller design problem into an LMI linear matrix inequality problem and use the CCL cone complement linearization algorithm to obtain a set of feasible PID controller gains K. P CCL , K I CCL , K D CCL and the feasible solution of the observer gain L CCL , and then the population initialization range of the GA genetic algorithm can be obtained; S32, combining GA genetic algorithm and LMI linear matrix inequality to obtain the observer gain L and PID controller gain to be designed, that is, the corresponding gains K of the three parts of the controller P , K I , K D ; S33, the UAV system state estimation vector given by the observer As input, the output ζ of the PID controller is obtained k .

4. The method for designing a secure PID controller assisted by an evolutionary algorithm considering network hybrid attacks according to claim 3 is characterized in that: Step S31 specifically includes the following steps: S311. Apply Schur's complement lemma to the matrix inequality with nonlinear terms of the PID controller, and obtain the corresponding matrix inequality as follows: in, I5=diag{I,I,I,I,I},Y=P -1 ; Among them, δ∈[0,1) is a known scalar, ρ is a given constant, Q = diag{Q1, Q2, ..., Q q }, P, Q, R are the positive definite matrices to be determined, I is the identity matrix, τ=0, λ is a given constant greater than 0, and γ is the performance index H to be determined. ∞ The value of Among them, A, B, C, D, E are parameter matrices, and T is a real matrix of known dimension; S312: Apply the CCL cone complement linearization algorithm to the linear matrix inequality obtained in S311 and other linear matrix inequalities to obtain a set of feasible PID controller gains K P CCL , K I CCL , K D CCL and the feasible solution of the observer gain L CCL , and then obtain the population initialization range of the GA genetic algorithm.

5. The method for designing a secure PID controller assisted by an evolutionary algorithm considering network hybrid attacks according to claim 3 is characterized in that: Step S32 specifically includes the following steps: S321, GA genetic algorithm population initialization when each individual initialization range is in, Indicates the proportional coefficient corresponding to each gain, which is given by experience; S322, randomly generate an initial population from the initialization range obtained in S321; S323, using the LMI tool to solve the linear matrix inequality group for each individual, if there is a solution, calculate the corresponding fitness function value; if there is no solution, assign a value to the fitness value of the individual; S324, executing selection, crossover, mutation, and elite retention GA operations; S325. If the maximum number of iterations is reached, the optimal individual gain K obtained during the iteration is saved. P , K I , K D and observer gain L, otherwise return to step S323.

6. The method for designing a secure PID controller assisted by an evolutionary algorithm considering network hybrid attacks according to claim 3 is characterized in that: In step S33, the UAV system state estimation vector given by the observer is and the output of the PID controller ζ k The formula is: Among them, y k is the total output of the hybrid attack network model, represents the estimated state vector of the UAV system at time k, They represent the UAV system state estimation vector at time k+1 and time k-1 respectively, represents the accumulation of the UAV system state estimation vector from time kq to time k-1, all given by the observer, u k represents the control signal at the latest moment, and the parameter q is a given positive scalar; ζ k Represents the output of the PID controller; K P , K I , K D Represents the PID controller gain to be designed, corresponding to the three parts of proportion P, integration I, and differentiation D respectively, and L represents the observer gain to be designed.

7. The method for designing a secure PID controller assisted by an evolutionary algorithm considering network hybrid attacks according to claim 1, characterized in that: The event triggering mechanism adopted in step S4 is specifically: Design event trigger function f(Φ k ,δ), the formula is expressed as: Among them, δ∈[0,1) is a known scalar, Represents the output of the PID controller at the latest moment; when the trigger function f(Φ k ,δ)>0, when the event trigger mechanism is activated, the output of the PID controller at the latest moment is As the input of the actuator, the latest control signal u is obtained k , the formula is: When the event trigger mechanism is not in effect, the latest control signal remains unchanged from the control signal at the previous moment, that is, u k =u k-1 .

Images