An autonomous response control method, system and medium for an unmanned aerial vehicle

Abstract

The present invention provides an autonomous response control method, system and medium for unmanned aerial vehicles, belonging to the technical field of unmanned aerial vehicles, including: regarding the unmanned aerial vehicle cluster in the unmanned aerial vehicle system as multiple agents according to the communication topology relationship of the unmanned aerial vehicle system, and constructing a multi-agent system model under network attacks according to the attack type; when the attack type is a physical domain attack type, constructing an elastic control protocol based on the constraint index of meeting the energy consumption while ensuring safety; canceling the influence of the physical domain attack on the flight control of the unmanned aerial vehicle system through the elastic control protocol; when the attack type is an information domain attack type, constructing an adaptive elastic formation control protocol; regulating the states of current multiple agents through the adaptive elastic formation control protocol, and controlling the formation of the unmanned aerial vehicle system. This method eliminates or weakens the influence of attacks by changing the control protocol, so as to achieve safe control under attack conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles, and particularly to an autonomous response control method, system and medium for unmanned aerial vehicles. Background Art

[0002] As a high-dimensional complex system integrating information, physical and mission domains, the multi-domain system of unmanned aerial vehicles is inevitably subject to security threats such as complex environments on the ground, in the air, in the jungle, etc. and malicious network attacks when performing tasks such as fixed-point delivery, jungle traversal, cargo transportation, ultra-low altitude flight, penetration combat, etc., which will affect the safe flight of unmanned aerial vehicles.

[0003] The attack methods on the multi-domain system of unmanned aerial vehicles are diverse, including attacks on the physical domain and the information domain. In the physical domain, spoofing attacks and false data injection attacks are common means. For example, by spoofing GPS signals, attackers can mislead the navigation system of unmanned aerial vehicles and make them deviate from the predetermined flight path. False data injection attacks inject incorrect data into the sensor system of unmanned aerial vehicles, causing them to make incorrect judgments and reactions. These attacks can not only disrupt the normal operation of unmanned aerial vehicles, but also lead to serious safety accidents.

[0004] Existing technologies have many deficiencies in dealing with these multi-domain attacks. The lack of a rapid response mechanism within multiple domains is a major problem. When the unmanned aerial vehicle system is attacked, if the attack cannot be quickly identified and responded to, the impact of the attack often expands rapidly. Existing defense mechanisms are often static and lack dynamic adaptability. In a complex multi-domain attack environment, the means and strategies of attackers may change continuously, and static defense mechanisms are difficult to effectively cope with. Existing defense technologies also have deficiencies in autonomous response. The unmanned aerial vehicle system needs to be able to automatically make appropriate reactions when attacked, rather than relying entirely on manual intervention. Summary of the Invention

[0005] To solve the above problems, the present invention provides an autonomous response control method, system and medium for unmanned aerial vehicles. This method can achieve a defense mechanism that can quickly respond, dynamically adapt and has autonomous response ability in the face of complex and changeable attacks. By improving existing technologies and enhancing the comprehensive defense ability of the system, the safety and reliability of the multi-domain system of unmanned aerial vehicles can be effectively improved, and its normal operation can be ensured.

[0006] To achieve the above object, the present invention provides the following technical solutions.

[0007] An autonomous response control method for unmanned aerial vehicles, comprising the following steps:

[0008] According to the communication topology relationship of the UAV system, the UAV cluster in the UAV system is regarded as multiple agents, and a multi-agent system model under network attack is constructed according to the attack type; among them, in this model, one agent is designated as the leader, and the other agents are followers;

[0009] When the UAV system is under physical domain attack, a distributed attack estimation observer is designed according to the multi-agent system model, and the network attack is reconstructed based on the distributed attack estimation observer to obtain the estimated result of the attack; after obtaining the attack estimation result, on the premise of ensuring safety, based on the constraint index of energy consumption, an elastic control protocol is constructed to offset the influence of the physical domain attack on the flight control of the UAV system and ensure that the UAV can still maintain a stable flight state when under attack; among them, the types of the physical domain attack include spoofing attack and false data injection attack;

[0010] When the UAV system is under information domain attack, a formation function is defined to determine the set of information domain attacks suffered on each channel; the union of the time when an attack occurs on the attacked channel and the time when the non-attacked channel occurs is constructed; based on the union, an adaptive elastic formation control protocol is constructed to regulate the states of current multiple agents and control the formation of the UAV system; among them, the types of the information domain attack include denial-of-service attack and malware attack.

[0011] Preferably, constructing the multi-agent system model under network attack according to the attack type includes the following steps:

[0012] When the UAV system is under physical domain attack, the multi-agent system model is:

[0013] Leader:

[0014] ;

[0015] Followers:

[0016] ;

[0017] Among them, and are known system matrices, is the attack distribution matrix, is the physical domain network attack signal suffered on the th agent, is the leader dynamics model, is the follower dynamics model, represents the number of followers, and represent the state, output vector and Lipschitz nonlinear term of the leader respectively, and respectively represent the system state of the th follower, the control protocol to be designed, the system output, and the Lipschitz nonlinear term.

[0018] Preferably, when the UAV system is under physical domain attacks, design a distributed attack estimation observer according to the multi-agent system model, and reconstruct the cyber attack based on the distributed attack estimation observer to obtain the estimated result of the attack, including the following steps:

[0019] is the physical domain cyber attack signal suffered by the th agent, described as the following bounded and unknown nonlinear function:

[0020] ;

[0021] where is an unknown constant matrix, is an unknown time-varying function related to time t , state x and output y ;

[0022] Design the following state observer for the multi-agent system model:

[0023] ;

[0024] where and are known system matrices, is the attack distribution matrix, is the output of the th follower agent, and are respectively the state estimate and output estimate of the th follower agent, is the estimated value of the system state, is the control protocol, and are the output and output estimate of the initial follower agent, and are the output and output estimate of the agent neighbor, is the Lipschitz nonlinear term estimate, is the th agent neighbor set; is the interaction weight between the th agent and the leader, represents the and iThe interaction weights between agents; is the gain matrix of the observer, is the relative output estimation error of the th follower, and the attack estimation on the

[0025] th

[0026] agent follows the following rule: is the given adaptive learning rate, is a constant matrix, is the first derivative of with respect to time t, and is the first derivative of

[0027] Preferably, after obtaining the attack estimation result, on the premise of ensuring safety, an elastic control protocol is constructed based on the constraint index of energy consumption as follows:

[0028] ;

[0029] where, when is , otherwise ; is the relative output error, is the controller gain matrix, is the adaptive parameter, is called the control input signal, i.e., the formation protocol, respectively represent the system output of the leader, the th follower and the th follower's system output, represents a constant matrix that satisfies is the attack distribution matrix, is the attack estimation on the th agent; is the interaction weight between the th agent and the is a positive definite gain matrix, represents the transpose of the input signal ; and respectively represent the actual energy consumption and the upper bound of the given maximum energy constraint. The goal of the control is to meet the energy consumption constraint index while ensuring safety. Implement , and are respectively the state of the th agent and the state of the th agent.

[0030] Preferably, the multi-agent system model under network attack constructed according to the attack type includes the following steps:

[0031] When the UAV system is under information domain attack, the multi-agent system model is:

[0032] Considering 1 leader, followers, the following multi-agent system model is given:

[0033] ;

[0034] Among them, and represent the position and velocity vectors of the leader, is the agent label, and respectively represent the position vector and velocity vector of the th agent; represents the control protocol of the th agent, is a known constant; is the leader dynamics model, is the follower dynamics model; is the output of the leader agent, is the output of the follower agent.

[0035] Preferably, the definition of the formation function to determine the set of information domain attacks on each channel; constructing the union of the attack time on the attacked channel and the time on the non-attacked channel includes the following steps:

[0036] Define as a continuously differentiable formation function, as a formation function related to position, as a formation function related to velocity, and define the states of the leader and followers as The symbol represents the transpose of a vector or matrix, are respectively the error between the leader and the formation function and the error between the follower and the formation function, They are the errors of the two component vectors of position and velocity with respect to the formation function. Considering is 0, then we have:

[0037] ;

[0038] where and are and the first-order derivatives with respect to time t; , represents the Kronecker product, is the elastic formation control protocol to be designed; is the first-order derivative of the formation function with respect to time t;

[0039] Considering that multiple channels are simultaneously under a denial-of-service attack, define as the set of sequences from no attack to attack transition. Then:

[0040] ;

[0041] where represents the th time interval, is the duration of the attack in the th time interval; Considering , then:

[0042] ;

[0043] where and respectively represent the sets of time intervals of communication denial and communication connectivity within the time interval ; For the expression of , represents multiple attacked time periods within the time interval from to , that is, the union of these time periods constitutes , which is also the time period of communication denial; represents the time interval from to , excluding all attacked time intervals , that is, the time interval of communication connection;

[0044] Define as the set of channels suffering from a denial-of-service attack within the time period . Assume:

[0045] ;

[0046] Among them, , represents the attack intensity, , represents a denial-of-service attack; define the set F( t ) of channels under attack at time

[0047] ;

[0048] Among them, is the set of all edges; define the union of the time when an attack occurs on the attacked channels and the time when the non-attacked channels occur during the time period:

[0049] ;

[0050] Among them, is the time when an attack occurs on the attacked channels, is the time when the non-attacked channels occur.

[0051] Preferably, the adaptive elastic formation control protocol is:

[0052]

[0053] ;

[0054] Among them, is the control quantity, is the error between the leader and the formation function, and are the errors between the agents and agent as followers and the formation function; and are the gain matrices, is a known constant; represents the performance function of the entire multi-agent system; and are the controller gain matrix and the adaptive weight gain matrix to be designed; is the action weight of the leader and agent , is the adaptive change coefficient of the action weight of the leader and agent ; is from agent to agent The action weight; if the agent receives the information of agent j, , otherwise, ; is the action weight adaptive change coefficient from agent to agent , and its initial value , is the first-order derivative with respect to time t; is the follower formation function, and the core idea of the performance formation control protocol is to achieve the formation of multiple agents while determining the upper bound of the performance index ; is the first-order derivative with respect to time t of the formation function related to speed;

[0055] Obtained through derivation and calculation , and select appropriate adjustment parameters , and obtain the allowable attack intensity according to the following linear matrix inequality:

[0056] ;

[0057] wherein, is the attenuation rate, and the attack intensity are both scalars, refers to the set of all edges after removing the set of channels attacked at time .

[0058] An autonomous response control system for an unmanned aerial vehicle, the system includes:

[0059] A processor;

[0060] A memory, on which a computer program that can run on the processor is stored;

[0061] wherein, when the computer program is executed by the processor, the steps of the autonomous response control method for the unmanned aerial vehicle are implemented.

[0062] A computer-readable storage medium, on which a data processing program is stored, and when the data processing program is executed by a processor, the steps of the autonomous response control method for the unmanned aerial vehicle are implemented.

[0063] The beneficial effects of the present invention:

[0064] The present invention proposes a method, system, and medium for autonomous response control of unmanned aerial vehicles (UAVs). This method uses a distributed attack estimation observer to continuously monitor the states of various agents in the UAV system, identify abnormal behaviors, and reconstruct attack patterns. Based on the reconstructed attack estimation results, the system can quickly adjust the control strategy and adopt an elastic control protocol to counter attacks, ensuring the flight safety and stability of the UAVs. For information domain attack types such as denial-of-service (DoS) attacks and malware attacks, we define a formation function to determine the time distribution of attacked channels and non-attacked channels. The adaptive elastic formation control protocol can dynamically adjust the communication and cooperation methods among agents according to the attack situation of the channels, ensuring that the UAV swarm can still maintain an effective formation and coordination under information domain attacks. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is a flowchart of the method for autonomous response control of UAVs according to an embodiment of the present invention;

[0066] Figure 2 are the attack reconstruction simulation results of Agent 1 and Agent 2, where Figure 2 (a) thereof is the attack reconstruction simulation result of Agent 1; Figure 2 (b) thereof is the attack reconstruction simulation result of Agent 2;

[0067] Figure 3 are the agent attack reconstruction simulation results of Agent 3 and Agent 4, where Figure 3 (a) thereof is the attack reconstruction simulation result of Agent 3; Figure 3 (b) thereof is the attack reconstruction simulation result of Agent 4;

[0068] Figure 4 are the state errors between the leader and the follower of Agent 1 and Agent 2, where Figure 4 (a) thereof is the state error between the leader and the follower of Agent 1; Figure 4 (b) thereof is the state error between the leader and the follower of Agent 2;

[0069] Figure 5 are the state errors between the leader and the follower of Agent 3 and Agent 4, where Figure 5 (a) thereof is the state error between the leader and the follower of Agent 3; Figure 5 (b) thereof is the state error between the leader and the follower of Agent 4;

[0070] Figure 6 is the multi-agent topology structure;

[0071] Figure 7 is the multi-channel DoS attack signal;

[0072] Figure 8 For and the trajectories, where Figure 8 the (a) of X in the direction of the trajectory Figure 8 the (b) of Y in the direction of the trajectory;

[0073] Figure 9 For and the trajectories, where Figure 9 the (a) of Z in the direction of the trajectory Figure 9 the (b) of X in the direction of the trajectory;

[0074] Figure 10 For and the trajectories, where Figure 10 the (a) of Y in the direction of the trajectory Figure 10 the (b) of Z in the direction of the trajectory;

[0075] Figure 11 are the positions of the agents at different times;

[0076] Figure 12 are the performance function and the upper bound of the guaranteed performance. Specific implementation manners

[0077] Regarding the research on the autonomous, safe, fast response and iterative control of the UAV system, it is mainly based on a hierarchical and graded full-state index system, forms a fast response mechanism within the domain based on the combined optimization of the pre-plan library, constructs a real-time full-state feedback loop, and in the multi-domain threat state space, with multi-domain security as the overall goal, through layer-by-layer and step-by-step iterative optimization, realizes the full-state safe collaborative control of the UAV system. The specific research idea is to consider the UAV multi-domain system as a high-dimensional complex system that integrates the information, physical, and mission domains. When performing tasks such as fixed-point delivery, jungle crossing, cargo transportation, ultra-low altitude flight, penetration combat, etc., it is inevitably affected by security threats from complex environments such as the ground, air, jungle, and malicious network attacks, which will affect the safe flight of the UAV. Therefore, this embodiment proposes a method for autonomous response control of the UAV under information domain and physical domain attacks, as Figure 1 shown. Specifically, it includes the following steps:

[0078] According to the communication topology of the UAV system, the UAV clusters in the UAV system are regarded as multiple agents, and a multi-agent system model under cyber attacks is constructed according to the attack types. Among them, in this model, one agent is designated as the leader, and the other agents are followers.

[0079] When the UAV system is under physical domain attacks, a distributed attack estimation observer is designed according to the multi-agent system model, and the cyber attacks are reconstructed based on the distributed attack estimation observer to obtain the estimated results of the attacks. After obtaining the attack estimation results, on the premise of ensuring safety, an elastic control protocol is constructed based on the constraint index of energy consumption to offset the impact of physical domain attacks on the flight control of the UAV system and ensure that the UAV can still maintain a stable flight state when under attack. Among them, the types of the physical domain attacks include spoofing attacks and false data injection attacks.

[0080] When the UAV system is under information domain attacks, a formation function is defined to determine the set of information domain attacks suffered on each channel. The union of the time when attacks occur on the attacked channels and the time when non-attacked channels occur is constructed. Based on the union, an adaptive elastic formation control protocol is constructed to regulate the current states of each multi-agent and control the formation of the UAV system. Among them, the types of the information domain attacks include denial-of-service attacks and malware attacks.

[0081] Specifically, the UAV autonomous response control method under physical domain cyber attacks specifically includes the following steps:

[0082] S1: According to the communication topology of the UAV system, the UAV clusters in the UAV system are regarded as multiple agents, and a multi-agent system model under cyber attacks is constructed according to the attack types. Among them, in this model, one agent is designated as the leader, and the other agents are followers:

[0083] Leader:

[0084]

[0085] Follower:

[0086]

[0087] Among them, is the known system matrix, is the attack distribution matrix, is the physical domain cyber attack signal suffered on the th agent, is the leader dynamics model, is the follower dynamics model, represents the number of followers, and represent the state of the leader, the output vector, and the Lipschitz nonlinear term respectively, and represent the system state, the control protocol to be designed, the system output, and the Lipschitz nonlinear term of the -th follower respectively.

[0088] S2: Design a distributed attack estimation observer according to the multi-agent system model. Obtain the relative output estimation error through the attack estimation observer, reconstruct the suffered cyber-attack, and obtain the attack estimation result:

[0089] is the cyber-attack signal in the physical domain suffered by the -th agent, described as the following bounded and unknown nonlinear function:

[0090] .

[0091] where, is an unknown constant matrix, is an unknown time-varying function related to time , state and output ;

[0092] For the multi-agent system model, design the following state observer:

[0093] .

[0094] where, and are known system matrices, is the attack distribution matrix, is the output of the -th follower agent, and are the state estimation and output estimation of the -th follower agent respectively, is the system state estimation value, is the control protocol, and are the output and output estimation of the initial follower agent, and are the output and output estimation of the agent neighbors, is the Lipschitz nonlinear term estimation, is the set of neighbors of the -th agent; is the interaction weight between the -th agent and the leader, Represents the and i weights of interaction between agents; is the gain matrix of the observer, is the relative output estimation error of the th follower, and the attack estimation on the

[0095] th

[0096] agent follows the following rule: where is a given adaptive learning rate, is a constant matrix, is the first-order derivative of with respect to time t, and is the first-order derivative of

[0097] S3: After obtaining the attack estimation result, on the premise of ensuring safety, a resilient control protocol is constructed based on the constraint index of energy consumption to offset the impact of physical domain attacks on the flight control of the UAV system and ensure that the UAV can still maintain a stable flight state when under attack:

[0098]

[0099] where, when is true, , otherwise ; is the relative output error, is the controller gain matrix, is the adaptive parameter, is called the control input signal, i.e., the formation protocol, respectively represent the system output of the leader, the th follower and the th follower's system output, represents a constant matrix that satisfies is the attack distribution matrix, is the attack estimation on the th agent; is the weight of interaction between the th agent and the leader; is a positive definite gain matrix, represents the transpose of the input signal ​ and represent the actual energy consumption and the upper bound of the given maximum energy constraint respectively. The goal of the control is to meet the energy consumption constraint index while ensuring safety achieve , and are the state of the th agent and the state of the th agent respectively.

[0100] In this embodiment, the scheme takes a multi-agent system with 1 leader and 4 followers as an example for simulation verification, numbered and respectively. Each agent has a 4-dimensional state vector. It is assumed that the attack occurs on follower 1 and follower 4, and the attack signal and its estimate are as shown in Figure 2 and Figure 3 . Figure 2 (a) of Figure 2 (b) of Figure 3 (a) of Figure 3 (b) of

[0101] After the attack reconstruction, according to the designed control protocol, the impact of the attack on the system is offset to achieve elastic secure consensus control. The simulation results are as shown in Figure 4 and Figure 5 . Figure 4 (a) of Figure 4 (b) of Figure 5 (a) of Figure 5 (b) of

[0102] Specifically, the autonomous response control method of the UAV under the information domain network attack specifically includes the following steps:

[0103] S1: According to the communication topology relationship of the UAV system, regard the UAV cluster in the UAV system as multiple agents, and construct a multi-agent system model under network attack according to the attack type; one agent is the leader, and other agents are followers:

[0104] Considering 1 leader, followers, the following multi-agent system model is given:

[0105]

[0106] Among them, and represent the position and velocity vectors of the leader, is the agent label, respectively represent the position vector and velocity vector of the -th agent; represents the control protocol of the -th agent, is a known constant; is the leader dynamics model, is the follower dynamics model; is the output of the leader agent, is the output of the follower agent.

[0107] S2: When the UAV system is under an information domain attack, a formation function can be defined to determine the set of channels under information domain attack. Define as the formation function:

[0108] Define as a continuously differentiable formation function, as the position-related formation function, as the velocity-related formation function. Define the states of the leader and followers as The symbol represents the transpose of a vector or matrix, are respectively the error between the leader and the formation function and the error between the follower and the formation function, are respectively the errors of the position and velocity component vectors with the formation function. Considering is 0, then there is:

[0109]

[0110] Among them, and are and at the first derivative with respect to time t; , represents the Kronecker product, is the elastic formation control protocol to be designed; is the first derivative of the formation function with respect to time t;

[0111] Considering that multiple channels are simultaneously under a denial-of-service attack, define as the set of sequences for the transition from no attack to attack, then:

[0112] .

[0113] Among them, represents the th time interval, is the duration of the attack existence in the th time interval; Considering , then:

[0114] .

[0115] Among them, and respectively represent the sets of time intervals of communication rejection and communication connectivity within the time interval ; For the expression of , represents multiple attacked time periods within the time interval from to , that is, the union of these time periods constitutes , which is also the time period of communication rejection; represents the time interval from to , excluding all attacked time intervals , that is, the time interval of communication connection.

[0116] Define as the set of channels suffering from denial - of - service attacks on the time period, assuming: .

[0117] .

[0118] Among them, , represents the attack intensity, , represents the denial - of - service attack; Define the set F( ) of channels attacked at the t moment as:

[0119] .

[0120] Among them, is the set of all edges; Define the union of the time when attacked channels are attacked and the time when non - attacked channels occur within the time period:

[0121] .

[0122] Among them, is the time when attacked channels are attacked, is The occurrence time of the non-attacked channel.

[0123] S3: The adaptive elastic formation control protocol is:

[0124]

[0125] Where, is the control variable, is the error between the leader and the formation function, and are the agents and the agent as the error between the follower and the formation function; and are the gain matrices, are known constants; represents the performance function of the entire multi-agent system; and are the controller gain matrix and the adaptive weight gain matrix to be designed; is the action weight of the leader and the agent , is the adaptive change coefficient of the action weight of the leader and the agent ; is the agent to the agent 's action weight; if the agent receives the information of agent j, , otherwise, ; is the adaptive change coefficient of the action weight of the agent to the agent , whose initial value , is the first-order derivative of with respect to time t; is the follower formation function, and the core idea of the guaranteed performance formation control protocol is to achieve the formation of multi-agents while determining the upper bound of the performance index;

[0126] is obtained through derivation and calculation and a suitable is selected, according to the following linear matrix inequality:

[0127] ;

[0128] ;

[0129] 。

[0130] Obtain the allowable attack strength.

[0131] Among them, is the attenuation rate, and the attack strength are both scalars, refers to the set of all edges after removing the set of channels under attack at time .

[0132] S4: Regulate the states of current multi-agents through the adaptive elastic formation control protocol and control the formation of the UAV system.

[0133] In this embodiment, the feasibility of the scheme is verified through simulation. Taking 1 leader and 3 followers as an example, the connection form between agents and the forms of attacks suffered by each channel are as Figure 6 and Figure 7 shown. Under the designed elastic formation control protocol, the simulation results are as Figure 8 , Figure 9 and Figure 10 shown, which are the trajectories in the direction for and . Among them, Figure 8 (a) of X is the trajectory in the direction for Figure 8 (b) of Y is the trajectory in the direction for Figure 9 (a) of Z is the trajectory in the direction for Figure 9 (b) of X is the trajectory in the direction for Figure 10 (a) of Y is the trajectory in the direction for Figure 10 (b) of Z is the trajectory in the direction. It can be seen from Figure 8 , Figure 9 and Figure 10 that the related to position and velocity all finally converge to the state of the leader, that is, a safe formation is achieved.

[0134] Figure 11 represents the positions of the agents in the three-dimensional space when forming formations at different times, andFigure 12 It shows that the representative is the performance and its upper bound. The left axis corresponds to the performance function, and the right axis corresponds to the performance upper bound. It can be seen from the figure that the performance function index is far less than its upper bound, that is, while ensuring safety, the formation performance is guaranteed.

[0135] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for autonomous response control of a UAV, characterized in that: The following steps are involved: According to the communication topology of the UAV system, the UAV cluster in the UAV system is regarded as multiple agents, and a multi-agent system model under network attack is constructed according to the attack type; in this model, one agent is designated as the leader and the other agents are followers; When the UAV system is attacked by the physical domain, a distributed attack estimation observer is designed according to the multi-agent system model, and the network attack is reconstructed based on the distributed attack estimation observer to obtain the estimated result of the attack; after obtaining the attack estimation result, a flexible control protocol is constructed based on the constraint index of energy consumption under the premise of ensuring safety to offset the impact of the physical domain attack on the flight control of the UAV system and ensure that the UAV can still maintain a stable flight state when attacked; wherein the types of physical domain attacks include deception attacks and false data injection attacks; When the UAV system is attacked by the information domain, a formation function is defined to determine the set of channels attacked by the information domain; the union of the attack time of the attacked channel and the attack time of the unattacked channel is constructed; based on the union, an adaptive elastic formation control protocol is constructed to regulate the current state of each multi-agent and control the UAV system formation; wherein the types of information domain attacks include denial of service attacks and malware attacks; The method of constructing a multi-agent system model under network attack according to the attack type includes the following steps: When the UAV system is attacked in the physical domain, the multi-agent system model is: Leaders: Followers: in, and is the known system matrix, is the attack distribution matrix, is the physical domain network attack signal suffered by the k-th agent, For the leader dynamics model, is the follower dynamics model, N represents the number of followers, and g(x0(t)) represent the leader’s state, output vector, and Lipschitz nonlinear term, respectively. and g(x k (t)) represent the system state of the k-th follower, the control protocol to be designed, the system output and the Lipschitz nonlinear term, respectively; When the UAV system is attacked in the physical domain, a distributed attack estimation observer is designed according to the multi-agent system model, and the network attack is reconstructed based on the distributed attack estimation observer to obtain an estimated result of the attack, including the following steps: is the physical domain cyber attack signal suffered by the kth agent, which is described as the following bounded and unknown nonlinear function: f k (t)=M f Ψ(t,x,y); in, is the unknown constant matrix, is an unknown time-varying function related to time t, state x, and output y; For the multi-agent system model, the following state observer is designed: in, and is the known system matrix, is the attack distribution matrix, y k (t) is the output of the kth follower agent, and are the state estimate and output estimate of the kth follower agent, is the estimated value of the system state, For the control protocol, y0(t) and are the output and output estimate of the initial follower agent, y i (t) and are the outputs and output estimates of the agent’s neighbors, is the Lipschitz nonlinear term estimate, N k is the set of neighbors of the kth agent; ω k0 is the interaction weight between the kth agent and the leader, ω ki Represents the action weight between the kth and ith agents; is the observer gain matrix, is the relative output estimation error of the kth follower, The attack estimate on the kth agent follows the following rules: Among them, Γ is the given adaptive learning rate, F is the constant matrix, For k (t) is the first-order derivative at time t, for The first derivative at time t; After obtaining the attack estimation result, under the premise of ensuring safety, a flexible control protocol is constructed based on the constraint index of energy consumption, as shown below: Among them, when z k = 0, s k (t)=0, otherwise z k is the relative output error, K is the controller gain matrix, ρ is the adaptive parameter, u k is called the control input signal, i.e., the formation protocol, y 0, y i, y k represent the system output of the leader, the system output of the i-th follower, and the system output of the k-th follower, respectively. * Represents a satisfaction The constant matrix of is the attack distribution matrix, is the attack estimate on the kth agent; ω k0 is the interaction weight between the kth agent and the leader; ω ki is the interaction weight between the kth agent and the ith agent, H is a positive gain matrix, u k T Represents the input signal u k The transpose of J e and Represent the actual energy consumption and the given upper bound of the maximum energy constraint. The control goal is to meet the energy consumption constraint index under the premise of ensuring safety. accomplish x k and x i They are the k-th agent state and the i-th agent state respectively; The method of constructing a multi-agent system model under network attack according to the attack type includes the following steps: When the UAV system is attacked by the information domain, the multi-agent system model is: Consider 1 leader and N followers, given the following multi-agent system model: Where p0(t) and v0(t) represent the position and velocity vector of the leader, i=1,2,…N is the agent number, and Represent the position vector and velocity vector of the ith agent respectively; u i (t) represents the control protocol of the ith agent, α p , α v is a known constant; For the leader dynamics model, is the follower dynamics model; is the output of the leader agent, is the output of the follower agent; The formation function is defined to determine the set of channels that are attacked by the information domain; and the union of the attack time of the attacked channel and the attack time of the non-attacked channel is constructed, including the following steps: definition is a continuously differentiable formation function, f ip is the position-dependent formation function, f iv is the speed-dependent formation function, and the states of the leader and follower are defined as The symbol T represents the transpose of a vector or matrix. are the error between the leader and the formation function and the error between the follower and the formation function, ψ ip ,ψ iv are the errors between the position and velocity vectors and the formation function. Considering f0 is 0, we have: in, and is ψ0 and ψ i The first derivative at time t; represents the Kronecker product, It is a flexible formation control protocol to be designed; is the first-order derivative of the formation function at time t; Consider multiple channels being attacked by denial of service at the same time, define As a set of sequences from no attack to attack, then: H κ ={h κ }∪[h κ ,h κ +s κ ); Among them, h κ represents the kth time interval, s κ >0 is the duration of the attack at the kth time interval; considering t>s>0, then: in, and Θ(s,t) respectively represent the set of time intervals of communication rejection and communication connectivity in the time interval [s,t]; The expression of H κ ∩[s,t] represents multiple attacked time periods within the time interval from s to t, that is, the union of these time periods constitutes It is also the time period during which communication is denied; Θ(s,t) represents the time interval from s to t, excluding all time intervals under attack That is, the time interval of the communication connection; definition As the set of denial-of-service attacks on channel (i, j) in the time period [s, t), assume that: Among them, 0<μ ij <1,μ ij represents the attack strength, ξ ij >0,ξ ij represents a denial of service attack; the set of channels attacked at time t is defined as: Where E is the set of all edges; define the union of the attack time of the attacked channel and the attack time of the non-attacked channel in the time period [t1, t2]: in, [t1, t2] is the time when the attacked channel is attacked, Θ (i,j) (t1, t2) is the time when [t1, t2] is the channel that is not attacked; The adaptive elastic formation control protocol is: Among them, u i (t) is the control quantity, ψ0 is the error between the leader and the formation function, ψ i and ψ j is the error between agent i and agent j as followers and the formation function; α = [α p ,α v ]and is the gain matrix, α p , α v is a known constant; J e (t) represents the performance function of the entire multi-agent system; and is the controller gain matrix and adaptive weight gain matrix to be designed; ω i0,0 is the weight of the role of the leader and agent i, ω i0 (t) is the adaptive change coefficient of the weight of the leader and agent i; ω ij,0 is the weight of the action from agent j to agent i; if agent i receives information from agent j, ω ij,0 =1(i≠j), otherwise, ω ij,0 =0;ω ij (t) is the adaptive change coefficient of the action weight from agent j to agent i, and its initial value is ω ij (0) = ω ji (0) = 1, Yes ij (t) the first-order derivative with respect to time t; f i The core idea of ​​the performance formation control protocol is to determine the upper bound of the performance index. At the same time, realize the formation of multiple agents; is the first derivative of the velocity-related formation function with respect to time t; K is obtained by deducing u ,K w ,Q,β F , and select appropriate adjustment parameters The permissible attack strength is obtained according to the following linear matrix inequality: Among them, β F is the attenuation rate, and attack strength μ ij Both are scalars. ε\F(t) refers to the set of all edges ε minus the set of channels attacked at time t F(t).

2. An autonomous response control system for a drone, characterized in that: The system comprises: processor; a memory having stored thereon a computer program executable on the processor; Wherein, when the computer program is executed by the processor, the steps of the autonomous response control method of the drone as claimed in claim 1 are implemented.

3. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a data processing program, and when the data processing program is executed by the processor, the steps of the autonomous response control method of the unmanned aerial vehicle as claimed in claim 1 are implemented.

Images