Attitude consistency safety optimization control method for unmanned aerial vehicle cluster

Abstract

The invention discloses an attitude consistency safety optimization control method for an unmanned aerial vehicle cluster, and relates to the technical field of unmanned aerial vehicle control, and the method comprises the steps: building an attitude dynamic model of an unmanned aerial vehicle suffering from a spoofing attack and a non-autonomous leader; constructing a topological communication network according to the cooperation-competition relationship between the unmanned aerial vehicles, defining a synchronization error, constructing a preset time constraint function to constrain the synchronization error, and establishing a coordinate conversion equation; designing an optimal cost function and a Hamilton-Jacobian-Bellman equation, constructing an execution-evaluation structure based on a reinforcement learning algorithm, designing an execution-evaluation network weight updating law and a parameter updating law, and obtaining an approximate optimal virtual controller and a self-adaptive preset time attitude consistency safety optimization controller through online learning; and determining a to-be-designed control gain through stability analysis. According to the method provided by the invention, the attitude consistency security optimization control of the unmanned aerial vehicle cluster can be realized under the condition of network attack.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle control, and in particular to a method for safely optimizing attitude consistency control of a cluster of unmanned aerial vehicles. Background Art

[0002] With the rapid development of automatic control technology, single UAVs have been replaced by UAV swarm collaborative operations due to their limited operating range and work efficiency. It should be pointed out that attitude consistency control of UAV swarms has gradually become a research hotspot as the basic research of their collaborative flight. However, most existing control methods only consider the cooperative relationship between UAV swarms, but ignore the attitude consistency control problem under the coexistence of cooperative-competitive relationships.

[0003] The communication network between drone swarms is open and vulnerable to cyber attacks, which seriously threatens network security and information credibility. In particular, deception attacks are highly concealed and have attracted much attention. Attackers inject attack signals similar to the original data to mislead the system into making wrong decisions and thus destroy the stability of the system. In addition, given the current level of artificial intelligence, drone swarm systems lack the decision-making ability to deal with emergencies alone, and fully autonomous drone swarm collaborative control cannot be fully realized. More importantly, the adaptive optimization control strategy based on reinforcement learning is an effective way to balance control performance and control cost, but the existing optimization scheme may not be able to solve the problem of attitude consistency control of drone swarms under deception attacks.

[0004] Therefore, there is an urgent need to invent a safe optimization control method for the attitude consistency of UAV clusters to solve the problem of optimal control of the attitude consistency of UAV clusters under deception network attacks. Summary of the invention

[0005] In view of the shortcomings of existing drone control technology, the present invention provides a posture consistency safety optimization control method for drone clusters. By constructing a human-in-the-loop control framework, human decision-making ability is integrated into the intelligent control algorithm to improve the reliability of the drone cluster system. The Nussbaum gain technology is introduced to solve the problem of unknown deception attack weights. An execution-judgment network learning structure and a preset time constraint function are constructed, and an adaptive preset time posture consistency safety optimization controller based on reinforcement learning is designed to achieve posture consistency control of drone clusters with minimal control cost.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A method for controlling attitude consistency and safety optimization of a drone cluster, comprising:

[0008] Construct a dynamics model of the attitude system of a UAV and a non-autonomous leader that is subjected to a deception attack, wherein the dynamics model of the attitude system of the non-autonomous leader has the expected trajectory output by the non-autonomous leader and the input instructions given by the operator;

[0009] The synchronization error is defined based on the desired trajectory output by the non-autonomous leader and the topological communication network of the UAV swarm;

[0010] Construct a preset time constraint function, define auxiliary variables in combination with synchronization error, construct a composite function based on the auxiliary variables, calculate the conversion error based on the synchronization error and the composite function, and establish a coordinate conversion equation based on the conversion error based on the adaptive backstepping control technology;

[0011] According to the coordinate transformation equation, the first optimal cost function and the first Hamilton-Jacobi-Bellman equation are constructed, and the first Hamilton-Jacobi-Bellman equation is solved to obtain the optimal virtual control law. Introducing the first Nussbaum function and designing the parameter update law Execution-criticism network weight update law and Online adjustment and optimization to obtain a nearly optimal virtual controller

[0012] According to the coordinate transformation equation, the second optimal cost function and the second Hamilton-Jacobi-Bellman equation are constructed, and the second Hamilton-Jacobi-Bellman equation is solved to obtain the optimal actual control law. Introducing the second Nussbaum function and designing the parameter update law of the disturbance compensation signal Realize the estimation and compensation of identification error and external disturbance, and design parameter update law Execution-criticism network weight update law and The online optimization obtains the approximate optimal actual controller Λ il , denoted as the adaptive preset time posture consistency safety optimization controller;

[0013] The overall Lyapunov function is constructed, the stability of the closed-loop system is analyzed based on the Lyapunov stability theory, the control gain to be designed is determined, and the safe optimization control of the attitude consistency of the UAV cluster at a preset time is realized.

[0014] As a further description of the above technical solution:

[0015] Synchronization error E il for:

[0016]

[0017] Among them, sgn is the sign function, is the neighbor set of the i-th drone, Φ={1,2,...,N}, N represents the number of followers in the drone cluster, The i-th UAV can receive the k-th UAV information, and a is ik ≠0, where a ik >0 means that the i-th UAV and the k-th UAV are in a cooperative relationship, a ik <0 means that the i-th UAV and the k-th UAV are in a competitive relationship; It represents the damaged state signal of the i-th drone after being attacked by spoofing, represents the damaged state signal of the kth drone after being attacked by a spoofing attack, represents the expected trajectory of damage after the non-autonomous leader suffers a deception attack, where ε il (t), ε kl (t) and ε 0l (t) represents the deceptive attack signal injected by the malicious attacker, and Y 0l represents the system signals of the i-th UAV and the k-th UAV and the expected trajectory output by the non-autonomous leader; the i-th UAV can receive the leader information and is denoted as b i ≠0, otherwise recorded as b i =0, where b i >0 means that the i-th UAV has a cooperative relationship with the leader, b i <0 means that the i-th UAV is in a competitive relationship with the leader.

[0018] As a further description of the above technical solution:

[0019] Preset time constraint function for:

[0020]

[0021] Among them, α il and il is a positive design parameter that adjusts the convergence rate of the function, T il and c il To preset the convergence time and convergence accuracy.

[0022] As a further description of the above technical solution:

[0023] Composite Functions for:

[0024]

[0025] in, is a positive design parameter, r is the order of the UAV attitude system, is an auxiliary variable.

[0026] As a further description of the above technical solution:

[0027] Parameter update law for:

[0028]

[0029] Among them, γ il1 is the first error surface,

[0030] E il is the synchronization error, is the basis function, and κ il1 is the positive parameter to be designed;

[0031] Execution-criticism network weight update law and They are:

[0032]

[0033] in, and are learning rates, is the basis function;

[0034] Nearly Optimal Virtual Controller for:

[0035]

[0036] in, is the in-degree of node i, is the first Nussbaum function, k il1 and ∈ il1 is a positive design parameter,

[0037] As a further description of the above technical solution:

[0038] Parameter update law and They are:

[0039]

[0040] Among them, ε il2 , and κ il2 is the positive parameter to be designed, γ il2 is the second error surface, Δ il2 is a positive design parameter, is the basis function;

[0041] Execution-criticism network weight update law and They are:

[0042]

[0043] in, and are learning rates, is the basis function;

[0044] Approximate optimal actual controller Λ il for:

[0045]

[0046] in, is the second Nussbaum function, k il2 and ∈ il2 is a positive design parameter, Ψ il is the system parameter of the i-th UAV.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] (1) The method provided by the present invention introduces a human-in-the-loop control strategy in the adaptive backstepping control framework. The non-autonomous leader output instructions are adjusted in real time by the operator based on the flight conditions and mission requirements. The human decision-making ability and intelligent control algorithm are combined to improve the reliability and safety of the UAV cluster system. A coordinate transformation scheme including the output trajectory signal and attack weight of the attacked non-autonomous leader is constructed, which solves the design problem that the original correct state cannot be used for feedback control. The Nussbaum gain technology is introduced to effectively solve the problem of unknown deception attack weight, thereby ensuring the integrity and authenticity of the UAV cluster system information and maintaining network security.

[0049] (2) Different from the existing finite-time or fixed-time consistency control methods, the preset time constraint function constructed by the present invention realizes the synchronization error convergence constraint within the preset time, and removes the constraint of the traditional preset performance control method that the initial synchronization error must be within the finite initial performance boundary; in addition, the method provided by the present invention realizes the preset time attitude consistency safe optimal control of the UAV cluster with the minimum control cost by constructing a learning strategy based on the execution-judgment network structure, and under the invented optimization control algorithm, the tracking accuracy and convergence time can be flexibly adjusted according to the task requirements, thereby effectively improving the attitude consistency tracking performance of the UAV cluster.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the embodiments of the present invention are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 It is a flow chart of a method for safely optimizing attitude consistency of a drone cluster according to the present invention;

[0053] Figure 2 Design a block diagram for the adaptive preset time attitude consistency safety optimization controller of UAV swarm;

[0054] Figure 3 This is the topological communication network structure diagram of the drone cluster;

[0055] Figure 4 This is the attitude response curve of the UAV cluster;

[0056] Figure 5 This is the trajectory diagram of the synchronization error of the drone cluster;

[0057] Figure 6 Execution-criticism network weight trajectory diagram for drone swarm;

[0058] Figure 7 Input trajectory graph for drone swarm control. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0060] Please refer to Figure 1 and Figure 2 The present invention provides a method for safely optimizing the attitude consistency of a drone cluster, comprising the following steps:

[0061] Step S1: Establish the dynamics model of the attitude system of the UAV and the non-autonomous leader under deception attack, as follows:

[0062] The dynamic model of the attitude system of the i-th UAV is constructed as:

[0063]

[0064] Where i = 1, 2, ..., N; N represents the number of followers in the drone cluster; φi ,θ i and ψ i represents the Euler angle of the i-th UAV; and is the angular velocity of the i-th UAV; and is the angular acceleration of the i-th UAV; Λ i,φ ,Λ i,θ and Λ i,ψ is the control torque; and is the moment of inertia, i,φ ,Ξ i,θ and i,ψ is the aerodynamic damping coefficient; and Represents an unknown external disturbance.

[0065] In order to facilitate subsequent derivation, the attitude system dynamics model (1) of the i-th UAV is transformed into:

[0066]

[0067] in,

[0068]

[0069]

[0070] By introducing the human-in-the-loop control strategy, the intelligent control algorithm is combined with human decision-making to deal with emergencies in complex environments and improve the reliability of the UAV swarm system. The dynamic model of the non-autonomous leader attitude system is:

[0071]

[0072] in,

[0073] The desired trajectory of output for non-autonomous leaders,

[0074] Input instructions given by the operator,

[0075] is a nonlinear function;

[0076] Assume that the i-th drone and the non-autonomous leader are subject to the following spoofing attack:

[0077]

[0078] in, and Signals an error state that is subject to a spoofing attack;

[0079] μ0(Y0,t)=ε0(t)Y0;

[0080] ε i (t) = diag{ε i1 (t),ε i2 (t),ε i3 (t)},

[0081] and ε0(t)=diag{ε 01 (t),ε 02 (t),ε 03 (t)} is a deceptive attack signal injected by a malicious attacker;

[0082]

[0083] Step S2: Establish a topological communication network based on the cooperation-competition relationship between the drones in the drone cluster, and define the synchronization error E based on the expected trajectory output by the non-autonomous leader and the topological communication network. il , specifically including:

[0084] Directed Graph represents the topological communication network between N drones, where Φ = {1, 2, ..., N}, and are the vertex set, directed edge set, and adjacency matrix, respectively. is the neighbor set of the i-th drone, It means that the i-th UAV can receive the k-th UAV information, and a is recorded ik ≠0, It means that the i-th drone cannot receive the k-th drone information, and a is recorded ik =0, where a ik >0 means that the i-th UAV and the k-th UAV are in a cooperative relationship, a ik <0 means that the i-th drone and the k-th drone are in a competitive relationship. The Laplace matrix of the directed graph G is defined as L = D-Θ, where is the in-degree of node i. If the vertex set Φ can be divided into set Φ1 and set Φ2, and Φ1∪Φ2=Φ, s∈{1,2},a ik ≥0, and Then the directed graph is a structural equilibrium graph.

[0085] Define the leader as 0, and define the extended directed graph as in

[0086] where b i ≠0 means that the i-th drone can receive the leader information, b i = 0 means that the i-th UAV cannot receive the leader information, b i >0 means that the i-th UAV has a cooperative relationship with the leader, b i <0 means that the i-th UAV is in a competitive relationship with the leader.

[0087] According to the output expected trajectory Y0 of the non-autonomous leader and the topological communication network, the following synchronization error E is defined: il :

[0088]

[0089] Wherein, sgn is the sign function; i = 1, 2, ..., N; l = 1, 2, 3.

[0090] Step S3: according to the synchronization error E il , construct a preset time constraint function By designing composite functions Get the conversion error β il (t), based on the adaptive backstepping control technology, the coordinate transformation equation is established, including:

[0091] Construct the following preset time constraint function To achieve the synchronization error E il The convergence constraints are:

[0092]

[0093] Among them, α il and il is a positive design parameter that adjusts the convergence rate of the function, T il and c il To preset the convergence time and convergence accuracy.

[0094] The preset time attitude consistent safety optimization control is achieved by ensuring the following inequality holds:

[0095]

[0096] Based on the above inequality constraints, define auxiliary variables and They represent the synchronization error E il To Constraint Boundary and According to the auxiliary variable Constructing composite functions for:

[0097]

[0098] in, and is a positive design parameter, and r is the order of the UAV attitude system.

[0099] Furthermore, according to the synchronization error E il and composite functions Calculate the conversion error β il (t) is:

[0100]

[0101] Based on the adaptive backstepping control technology, the following coordinate transformation equation is established:

[0102]

[0103] Among them, γ il1 and γ il2 are the first error surface and the second error surface respectively, is a nearly optimal virtual controller.

[0104] Step S4: construct a first optimal cost function according to the coordinate transformation equation and the first Hamilton-Jacobi-Bellman equation Solving the Hamilton-Jacobi-Bellman equation Obtaining the optimal virtual control law Design parameter update law and the execution-critic network weight update law and Online adjustment and optimization to obtain a nearly optimal virtual controller Specifically include:

[0105] In order to achieve the optimal and preset time convergence constraint goals, define the first optimal cost function

[0106]

[0107] in, Ω represents the admissible control set including the origin, is the optimal virtual control law, is a nearly optimal virtual controller.

[0108] Based on the Bellman optimality principle, the first Hamilton-Jacobi-Bellman equation Defined as:

[0109]

[0110] By solving Obtaining the optimal virtual control law for:

[0111]

[0112] in,

[0113]

[0114] because The unknown cannot be used to optimize the control objective, so it is decomposed into:

[0115]

[0116] Among them, k il1 and ∈ il1 is a positive design parameter, is the first Nussbaum function,

[0117] is an unknown function,

[0118]

[0119] According to formula (13) and formula (14), we can calculate:

[0120]

[0121] because For unknown functions, radial basis function neural network is used Online identification To satisfy:

[0122]

[0123] in, is the unknown weight vector, is the basis function, represents the input vector, is an approximate error.

[0124] Furthermore, a reinforcement learning algorithm based on an execution-critic network is used to achieve the posture consistency optimization control goal and construct a judgment network approximate We can get:

[0125]

[0126] in, and They are and The estimated value of is the basis function, Judging the network weight update law and parameter update law Designed for:

[0127]

[0128] Among them, κ il1 is the positive parameter to be designed, is the learning rate.

[0129] Constructing the execution network Online adjustment of optimal virtual control law Obtaining a near-optimal virtual controller

[0130]

[0131] in, for The estimated value of the network weight update law is executed Designed for:

[0132]

[0133] in, is the learning rate.

[0134] According to equations (12), (17) and (20), the first Hamilton-Jacobi-Bellman equation is is approximated as:

[0135]

[0136] in,

[0137] Define the Hamiltonian residual error Δ il1 for:

[0138]

[0139] By designing a near-optimal virtual controller Make Established. Since the solution of the Hamilton-Jacobi-Bellman equation is unique, then is equivalent to:

[0140]

[0141] Define a continuous positive definite function Formula (21) is equivalent to F il1 =0, due to Therefore Therefore, the execution of the design can be judged - the network update law and It can be guaranteed that formula (21) holds.

[0142] Step S5: Construct the second optimal cost function and the second Hamilton–Jacobi–Bellman equation Solving the second Hamilton-Jacobi-Bellman equation Obtain the optimal actual control law Design parameter update law of disturbance compensation signal Realize the estimation and compensation of identification errors and external disturbances; design parameter update law Execution-criticism network weight update law and The online optimization obtains the approximate optimal actual controller Λ il , denoted as the adaptive preset time posture consistency safety optimization controller, specifically including:

[0143] According to the coordinate transformation equation, the second optimal cost function is defined as follows:

[0144]

[0145] in, To minimize The optimal actual control law, Λ il is an approximate optimal virtual controller, denoted as an adaptive preset time posture consistency safety optimization controller.

[0146] Based on the Bellman optimality principle, the second Hamilton-Jacobi-Bellman equation Defined as:

[0147]

[0148] By solving Obtaining the optimal virtual control law for:

[0149]

[0150] in,

[0151] because Unknown, so decompose it into:

[0152]

[0153] in, and They are and The first derivative of is the second Nussbaum function,

[0154] According to formula (27) and formula (28), we can calculate:

[0155]

[0156] Among them, k il2 and ∈ il2 is a positive design parameter.

[0157] Radial Basis Function Neural Network Online identification of unknown functions We can get:

[0158]

[0159] in, is the unknown weight vector, Represents input The basis function of is an approximate error.

[0160] Similar to step S4, construct the judgment network approximate We can get:

[0161]

[0162] in, and for and The estimated value of and is the upper bound of the deception attack weight and unknown perturbation, Represents input The basis function, Δ il2 is a positive design parameter.

[0163] Parameter update law Parameter update law And judge the network weight update law Designed for:

[0164]

[0165] Among them, εil2 , and κ il2 is the positive parameter to be designed, is the learning rate.

[0166] Constructing the execution network Optimize the optimal actual control law and obtain the approximate optimal actual controller Λ il , denoted as the adaptive preset time posture consistency safety optimization controller:

[0167]

[0168] in, for The estimated value of the network weight update law is executed Designed for:

[0169]

[0170] in, is the learning rate.

[0171] Step S5: construct an overall Lyapunov function, analyze the stability of the closed-loop system based on the Lyapunov stability theory, determine the control gain to be designed, and realize the safe optimization control of the attitude consistency of the UAV cluster at a preset time, which specifically includes:

[0172] According to the coordinate transformation equation (10), the following Lyapunov function V is constructed:

[0173]

[0174] in,

[0175] Derivative (37) is substituted into the designed approximate optimal virtual controller: Adaptive preset time posture consistency safety optimization controller Λ il , the weight update law of the execution-criteria network and Parameter update law And the parameter update law How to design parameters to meet Then the following inequality holds:

[0176]

[0177] in,

[0178]

[0179] For the matrix The minimum eigenvalue of .

[0180] Based on Lyapunov stability theory, the attitude consistency safety optimization control method of the UAV cluster described in the present invention can ensure that the attitude system of the UAV cluster is semi-globally bounded and the synchronization error is within a preset time T il Converges to the interval It also ensures that when the drone cluster is subjected to a deception attack, posture consistent and secure optimized control can be achieved with minimal control cost.

[0181] The following simulation experiments are carried out in the simulation software - MATLAB R2020a / SIMULINK to illustrate in detail the effectiveness and feasibility of the posture consistent safety optimization control method designed by the present invention.

[0182] The parameters of the drone attitude system are selected as follows:

[0183] Ξ i,ψ =Ξ i,θ =Ξ i,φ =0.6kg / rad,

[0184] The external unknown disturbance settings are:

[0185] The initial conditions of the UAV and the parameter update law are selected as:

[0186] The controller parameters are chosen as:

[0187] α il =0.3,υ il =2, c il =0.3, T il =3, r=2,k 11j =25, k 12j =k 13j =k 2lj =k 3lj =k 4l1 =30,,k 4l2 =20,∈ ilj =2,κ ilj =2, ε il2 =1, Δ il2 =exp(-10t).

[0188] The Gaussian function is chosen as:

[0189]

[0190] The deception attack weight is selected as: ε0(t) = diag{e -2t -0.01,-0.02sin(t)-e -t , -0.2e -2t},

[0191]

[0192]

[0193] The attack signal is injected starting from t=5s.

[0194] The input command for a non-autonomous leader is:

[0195]

[0196] The simulation results are plotted on Figures 3 to 7 . Figure 3 The topological communication network structure diagram of the drone cluster is depicted in the figure, where L1 represents the non-autonomous leader, QUAV1-QUAV4 represent 4 followers, QUAV1 and QUAV3 are in a cooperative relationship, and are in a competitive relationship with QUAV2 and QUAV4. Figure 4 In the figure, sub-figures (a) to (c) respectively plot the expected output trajectory of the non-autonomous leader L1 and the output trajectory of the four followers. It can be seen from the figure that even if the non-autonomous leader and the attitude subsystem of the drone are subjected to deception attacks, the output trajectory of the four followers still achieves attitude consistency tracking control according to the given cooperation-competition relationship. Figure 5 In the figure, sub-figures (a) to (d) respectively give the synchronization error E il From the figure, it can be found that the invented adaptive preset time posture consistency safety optimization controller will synchronize the error E il At the preset time T il = Constrained to the expected steady-state error range within 3s In addition, the initial bound of the preset time constraint function is infinite, which completely removes the constraint on the initial error of the traditional preset performance method. Figure 6 In the figure, subgraphs (a) to (d) respectively plot the execution-criticism network weight trajectories of the four followers. It can be found that all the execution-criticism network weights are bounded. Figure 7 It is the control input curve trajectory diagram. Combined with the simulation experiment results, it can be concluded that the attitude consistency safety optimization control method disclosed in the present invention integrates human-in-the-loop control technology, Nussbaum gain technology, preset time constraint function and optimal control algorithm based on execution-judgment network structure, which ensures the preset time attitude consistency safety optimization control of the drone cluster under the condition of deception attack.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for safely optimizing attitude consistency of a drone cluster, characterized in that: include: Construct a dynamics model of the attitude system of a UAV and a non-autonomous leader that is subjected to a deception attack, wherein the dynamics model of the attitude system of the non-autonomous leader has the expected trajectory output by the non-autonomous leader and the input instructions given by the operator; The synchronization error is defined based on the desired trajectory output by the non-autonomous leader and the topological communication network of the UAV swarm; Construct a preset time constraint function, define auxiliary variables in combination with synchronization error, construct a composite function based on the auxiliary variables, calculate the conversion error based on the synchronization error and the composite function, and establish a coordinate conversion equation based on the conversion error based on the adaptive backstepping control technology; According to the coordinate transformation equation, the first optimal cost function and the first Hamilton-Jacobi-Bellman equation are constructed, and the first Hamilton-Jacobi-Bellman equation is solved to obtain the optimal virtual control law. Introducing the first Nussbaum function and designing the parameter update law Execution-criticism network weight update law and Online adjustment and optimization to obtain a nearly optimal virtual controller il ; According to the coordinate transformation equation, the second optimal cost function and the second Hamilton-Jacobi-Bellman equation are constructed, and the second Hamilton-Jacobi-Bellman equation is solved to obtain the optimal actual control law. Introducing the second Nussbaum function and designing the parameter update law of the disturbance compensation signal Realize the estimation and compensation of identification error and external disturbance, and design parameter update law Execution-criticism network weight update law and Online optimization obtains the approximate optimal actual controller il , denoted as the adaptive preset time posture consistency safety optimization controller; The overall Lyapunov function is constructed, the stability of the closed-loop system is analyzed based on the Lyapunov stability theory, the control gain to be designed is determined, and the safe optimization control of the attitude consistency of the UAV cluster at a preset time is realized.

2. The attitude consistency safety optimization control method of the UAV cluster according to claim 1 is characterized in that: Synchronization error E il for: Among them, sgn is the sign function, is the neighbor set of the i-th drone, {1,2,...,N}, where N represents the number of followers in the drone cluster, The i-th UAV can receive the k-th UAV information, and a is ik 0, where a ik 0 means that the i-th UAV and the k-th UAV are in a cooperative relationship, a ik <0 means that the i-th UAV and the k-th UAV are in a competitive relationship; It represents the damaged state signal of the i-th drone after being attacked by spoofing, represents the damaged state signal of the kth drone after being attacked by a spoofing attack, represents the expected trajectory of damage after the non-autonomous leader suffers a deception attack, where il (t) kl (t) and 0l (t) represents the deceptive attack signal injected by the malicious attacker, and Y 0l represents the system signals of the i-th UAV and the k-th UAV and the expected trajectory output by the non-autonomous leader; the i-th UAV can receive the leader information and is denoted as b i 0, otherwise recorded as b i 0, where b i 0 means that the i-th drone has a cooperative relationship with the leader, b i 0 means that the i-th UAV is in a competitive relationship with the leader.

3. The attitude consistency safety optimization control method of the UAV cluster according to claim 2 is characterized in that: Preset time constraint function for: Among them, α il and il is a positive design parameter that adjusts the convergence rate of the function, T il and c il To preset the convergence time and convergence accuracy.

4. The attitude consistency safety optimization control method of the UAV cluster according to claim 3 is characterized in that: Composite Functions for: in, is a positive design parameter, r is the order of the UAV attitude system, is an auxiliary variable.

5. The attitude consistency safety optimization control method of the UAV cluster according to claim 4 is characterized in that: Parameter update law for: in, il1 is the first error surface, E il is the synchronization error, is the basis function, and κ il1 is the positive parameter to be designed; Execution-criticism network weight update law and They are: in, and are learning rates, is the basis function; Approximate optimal virtual controller θ il for: Among them, l i =|b i |+d i , is the in-degree of node i, is the first Nussbaum function, k il1 and ∈ il1 is a positive design parameter, 6. The attitude consistency safety optimization control method of the UAV cluster according to claim 5 is characterized in that: Parameter update law and They are: in, il2 , and κ il2 is the positive parameter to be designed, γ il2 is the second error surface, Δ il2 is a positive design parameter, is the basis function; Execution-criticism network weight update law Hehe They are: in, and are learning rates, is the basis function; Approximate optimal actual controller Λ il for: in, is the second Nussbaum function, k il2 and ∈ il2 is a positive design parameter, Ψ il is the system parameter of the i-th UAV.

Images