A cluster unmanned aerial vehicle event-triggered fractional order adaptive resilience cooperative control method

Abstract

The application discloses a cluster unmanned aerial vehicle event-triggered fractional order adaptive resilience cooperative control method. First, considering the constraint of limited communication resources, an event-triggered communication mechanism is constructed, which reduces the communication frequency between the cluster unmanned aerial vehicles, and a reference signal estimator is designed based on the event-triggered communication mechanism, which estimates the reference instruction signal of the unmanned aerial vehicle by using the non-continuous neighbor state information. Secondly, an unknown state observer is designed to estimate the unknown state information after the deception attack, and an unknown term caused by the fault is estimated by combining an adaptive law and a neural network method. Finally, a fractional order sliding mode control method is designed to realize the tracking of the estimated state information to the reference signal. The application can be applied to the resilience cooperative control of the cluster unmanned aerial vehicle formation under the condition of limited communication.

Description

Technical Field

[0001] This invention addresses the problems of spoofing attacks and actuator failures in swarmed UAVs under limited communication bandwidth by designing an event-triggered fractional-order adaptive resilient cooperative control method for swarmed UAVs, which belongs to the field of resilient cooperative control. Background Technology

[0002] In recent years, unmanned aerial vehicles (UAVs) have seen widespread development in both civilian and military fields due to their flexibility, low cost, and safe operation. However, a single UAV has limited range and low payload capacity, making it difficult to cope with complex scenarios. In contrast, swarm UAVs, through multi-UAV collaboration, can significantly improve system payload capacity and mission execution efficiency. However, with the technological development and scale expansion of swarm UAVs, some complex problems have gradually emerged. The communication bandwidth of swarm UAVs is often limited, and when a large amount of communication information is transmitted simultaneously, it may lead to problems such as data packet loss or delays. In addition, swarm systems may suffer from actuator failures, which, as a common type of failure, can greatly affect the flight performance of UAVs and, in severe cases, even cause them to crash. In some special environments, swarm UAVs may also be affected by spoofing attacks, which can tamper with the data received by UAVs, thereby disrupting their collaboration. Therefore, designing a resilient control scheme to ensure the flight safety of swarm UAV systems under the constraint of limited communication bandwidth is of paramount importance. Summary of the Invention

[0003] This invention addresses the problem of actuator failure and spoofing attacks encountered by swarm drones under limited bandwidth conditions. It designs an event-triggered fractional-order adaptive resilient cooperative control method for swarm drones to ensure that swarm drones can still fly safely when encountering actuator failure and spoofing attacks.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A method for event-triggered fractional-order adaptive resilience cooperative control of swarmed unmanned aerial vehicles (UAVs) includes the following steps:

[0006] Step 1: Establish the UAV dynamics model;

[0007] Step 2: Based on the UAV dynamics model established in Step 1, construct the UAV dynamics model and actuator failure model under deception attack.

[0008] Step 3: Design an event-triggered communication mechanism, establish a reference signal estimator, and estimate the UAV reference command signal;

[0009] Step 4: Establish an unknown state observer to estimate the unknown state after a deception attack, and combine adaptive laws and neural network methods to learn the unknown terms caused by actuator failure.

[0010] Step 5: Based on the reference command estimation information obtained in Step 3 and the unknown state estimation information obtained in Step 4, design a fractional sliding mode control method, construct a resilient control law, and realize the tracking of the reference signal.

[0011] Furthermore, the UAV dynamics model established in step one is as follows:

[0012] Set up a swarm of drones consisting of N followers and 1 leader, where the i-th drone has the following dynamics:

[0013] (1)

[0014] in, , and Let be the three-dimensional coordinates of the i-th UAV. express First derivative, express First derivative, express First derivative; For the speed of the drone, Represents the heading angle. Representing the flight path angle, the derivative is expressed as:

[0015] (2)

[0016] in, , , and The thrust, drag, lift, and lateral force of the drone are measured separately. It's an attack angle. It is the tilt angle. It is the sideslip angle; It's about the quality of the drone. It is the coefficient of gravitational acceleration;

[0017] The thrust, drag, lift, and lateral force of a drone are expressed as follows:

[0018] (3)

[0019] in, It's the throttle opening. For maximum thrust, , It is air density. Wing area; , and These are the lift coefficient, drag coefficient, and lateral force coefficient, respectively, expressed by the following formulas:

[0020] (4)

[0021] in, , and It is a constant. , and To attack angle Correlation coefficient To match the sideslip angle Correlation coefficient;

[0022] definition Input signal vector to the drone, Let the UAV's position vector be... For the drone's velocity vector, , , Let be the derivatives of the corresponding signs in the UAV position vector, respectively. Simplifying equations (1) and (2), we get:

[0023] (5)

[0024] in, and The transformation matrix is ​​represented as follows:

[0025] (6).

[0026] Furthermore, step two specifically involves:

[0027] Considering the types of actuator failures faced by UAVs, the actuator failure model is defined as follows:

[0028] (7)

[0029] in, , It is the efficiency loss factor, j=1, 2, 3. This represents a deviation fault. The input signal deviation is caused by a fault, j=1, 2, 3. It is the UAV control input command signal; combined with equation (5), the UAV model under actuator failure is defined as:

[0030] (8)

[0031] in, The unknown item is caused by an actuator malfunction. It is the identity matrix;

[0032] The drone is affected by a deception attack, resulting in unreliable position and velocity status information. The measurement signal of the drone after the deception attack is represented as follows:

[0033] (9)

[0034] in, and Unknown signals injected into deception attacks and Known position and velocity signals measured for the UAV; UAV position vector affected by deception attacks. and velocity vector It becomes an unknown signal; based on the UAV model (8) under actuator failure, combined with a deception attack, the following UAV model is obtained:

[0035] (10)

[0036] in, It is a known velocity signal measured by the drone. The known vector of influence, It is an unknown vector affected by a deception attack.

[0037] Furthermore, step three specifically involves:

[0038] The cooperative bias of drones is defined as follows:

[0039] (11)

[0040] in, It is the communication weight between the i-th drone and the j-th drone. It is the communication weight between the i-th drone and the leader drone. This represents the distance between the i-th drone and the j-th drone. This represents the distance between the i-th drone and the leader drone. For the location information of the j-th UAV, Location information for the leader's drone;

[0041] By modifying (11), the reference command signal for the i-th UAV is obtained. for:

[0042] (12)

[0043] Design an event-triggered communication mechanism: The i-th drone can only obtain the status information of its neighboring drones when an event is triggered; the first trigger time is set to... Design the following event to trigger the update condition for communication:

[0044] (13)

[0045] in, Representing the current moment, Indicates the trigger time The next triggering moment, Denotes the infimum of a function. For the design of positive constants, The velocity information is estimated by the unknown state observer;

[0046] Due to deception attacks, the drone's own measured position and velocity signals... and speed signal Unreliable, using position and velocity information estimated by the observer in an unknown state. and As a transmitted signal;

[0047] Design a reference signal estimator to obtain continuous state information:

[0048] (14)

[0049] in, and The reference position and velocity information estimated by the reference signal estimator. It is a positive constant in the design and must satisfy the condition. ,in , Represents the minimum value of the set; for The symbolic function vector, for The symbolic function vector, and Represented as:

[0050] (15)

[0051] in, and Is the i-th drone at the trigger time The received position and velocity information of the j-th UAV; and Represents the i-th drone at the trigger time Received information on the location and speed of the leader drone.

[0052] Furthermore, the unknown state observer, adaptive law, and neural network established in step four are as follows:

[0053] Construct the following unknown state observer to estimate the true state information:

[0054] (16)

[0055] in, and For position and velocity information estimated by the observer in the unknown state, and To estimate information for deceiving attack signals, It is an unknown item The estimated information, and For the designed diagonal positive matrix, For the positive parameters of the design, m=1, 2, n=1, 2, 3; The optimal weight matrix for the neural network The estimated matrix, For Gaussian function vectors, The input signal is the Gaussian function vector, which the neural network uses to learn the function in (10). .

[0056] Furthermore, neural network learning (10) is used. , is represented as:

[0057] (17)

[0058] in, for A real matrix of dimension , representing the optimal weight matrix. for A dimensional Gaussian function vector, For the input vector, To achieve the optimal learning error, Represents the number of nodes in the neural network; Represented as:

[0059] (18)

[0060] in, Represents the exponent Power of 1 For the center value vector, Width;

[0061] The estimated value in (16) is updated using an adaptive law, specifically as follows:

[0062] (19)

[0063] in, It is a positive constant for the design.

[0064] Furthermore, the fractional-order sliding mode and toughness control law in step five are as follows:

[0065] Set the tracking error of the UAV state estimation signal to the estimated reference signal as follows: Based on this tracking error, the fractional-order sliding surface is designed as follows:

[0066] (20)

[0067] in, For constructing fractional-order sliding surfaces, For positive odd numbers, For a positive constant of design, for fractional derivative, It is a fractional operator; The definition of is:

[0068] (twenty one)

[0069] in, For fractional operators, express The nth derivative, It is the Gamma function.

[0070] Based on the fractional-order sliding surface, the following toughness control law is designed:

[0071] (twenty two)

[0072] in, and For the design of the positive parameter matrix, For the positive parameters of the design, m=1, 2, n=1, 2, 3; , for The absolute value of j, j=1, 2, 3; for The symbolic function vector, The integral symbol is used. express The second derivative vector.

[0073] The present invention has the following beneficial effects:

[0074] (1) In view of the problem of actuator failure and deception attack encountered by swarm UAVs, the present invention designs a swarm UAV resilient collaborative control method, which realizes safe collaborative flight of swarm UAVs under deception attack and actuator failure.

[0075] (2) The present invention adopts an event-triggered mechanism to replace the continuous time-triggered communication of traditional clustered drones, thereby reducing the number of communication times of the cluster system.

[0076] (3) The present invention designs an unknown state observer to estimate the true state of the UAV after a deception attack, and constructs a fault compensation method to compensate for actuator faults. Attached Figure Description

[0077] Figure 1 This is a topology diagram of a swarm of drones;

[0078] Figure 2 This is a control architecture diagram for a cluster of drones.

[0079] Figure 3 This is a flight trajectory diagram of a swarm of drones;

[0080] Figure 4 A location information map of the swarm of drones;

[0081] Figure 5 A cooperative deviation diagram of the follower drone;

[0082] Figure 6 Estimate the bias map for the unknown state observer of the follower drone;

[0083] Figure 7 A tracking error diagram for the follower drone;

[0084] Figure 8 A diagram showing the actual control input signals for the follower drone;

[0085] Figure 9 This is a graph showing the number of communications by the follower drone. Detailed Implementation

[0086] The control method of the present invention will be further explained in conjunction with the accompanying drawings and tables.

[0087] This application proposes an event-triggered fractional-order adaptive resilient cooperative control method for swarmed UAVs. First, a dynamic model of the UAVs is constructed. To address spoofing attacks and actuator failures, the dynamic model is modified to obtain UAV dynamic models under these conditions. Second, an event-triggered communication mechanism is designed, and based on this mechanism, a reference signal estimator is built to reduce the number of communications between swarmed UAVs and to estimate the UAV reference signal. Then, based on neural networks and adaptive methods, an unknown state observer is designed to estimate the true state after a spoofing attack and compensate for the actuator failure. Finally, combining fractional-order calculus, a fractional-order sliding surface is designed, and a resilient control law for the UAVs is constructed based on this sliding surface to achieve reference signal tracking.

[0088] (a) Establishing the dynamic model of the UAV

[0089] Set up a swarm of drones consisting of N followers and 1 leader, where the i-th drone has the following dynamics:

[0090] (1)

[0091] in, , and Let be the three-dimensional coordinates of the i-th UAV. express First derivative, express First derivative, express First derivative. For the speed of the drone, Represents the heading angle. Representing the flight path angle, its derivative can be expressed as:

[0092] (2)

[0093] in, , , and The thrust, drag, lift, and lateral force of the drone are measured separately. It's an attack angle. It is the tilt angle. It is the sideslip angle; It's about the quality of the drone. This is the coefficient of gravitational acceleration. Thrust, drag, lift, and lateral force are expressed as:

[0094] (3)

[0095] in, It's the throttle opening. For maximum thrust, , It is air density. Wing area; , and These are the lift coefficient, drag coefficient, and lateral force coefficient, respectively, expressed by the following formulas:

[0096] (4)

[0097] in, , and It is a constant. , and To attack angle Correlation coefficient To match the sideslip angle Correlation coefficient;

[0098] definition Input signal vector to the drone, Let the UAV's position vector be... Given the UAV's velocity vector, equations (1) and (2) can be simplified to obtain:

[0099] (5)

[0100] in, and The transformation matrix is ​​represented as follows:

[0101] (6)

[0102] (b) Establishing a deception attack model and an actuator failure model

[0103] Considering actuator failures faced by drones, the failure model is represented as follows:

[0104] (7)

[0105] in, , It is the efficiency loss factor, j=1, 2, 3. This represents a deviation fault. The input signal deviation is caused by a fault, j=1, 2, 3. It is the UAV control input command signal; combined with (5), the UAV model under actuator failure is:

[0106] (8)

[0107] in, The unknown item is caused by an actuator malfunction. It is an identity matrix.

[0108] The drone was subjected to a spoofing attack, resulting in unreliable position and velocity status information. The drone's measurement signals after the spoofing attack are represented as follows:

[0109] (9)

[0110] in, and Unknown signals injected into deception attacks and Known position and velocity signals measured for the drone. Due to the impact of spoofing attacks, and It becomes an unknown signal. Based on the UAV model (8) with actuator failure, and further considering deception attacks, the following UAV model is obtained:

[0111] (10)

[0112] in, Is the recipient The known vector of influence, It is an unknown vector affected by a deception attack.

[0113] (c) Design an event-triggered communication mechanism and establish a reference signal estimation mechanism.

[0114] The cooperative bias of drones is defined as follows:

[0115] (11)

[0116] in, It is the communication weight between the i-th drone and the j-th drone. It is the communication weight between the i-th drone and the leader drone. This represents the distance between the i-th drone and the j-th drone. This represents the distance between the i-th drone and the leader drone. For the location information of the j-th UAV, The location information of the leader drone. By changing (11), the reference command signal of the i-th drone can be obtained. for:

[0117] (12)

[0118] Design an event-triggered communication mechanism: The i-th drone can only obtain the status information of its neighboring drones when an event is triggered; the first trigger time is set to... Design the following event to trigger the update condition for communication:

[0119] (13)

[0120] in, Representing the current moment, Indicates the trigger time The next triggering moment, Denotes the infimum of a function. For a positive constant of design, The velocity information is estimated by the unknown state observer.

[0121] Due to the existence of deception attacks, the information measured by the drone itself... and Unreliable; therefore, the position and velocity information estimated by the unknown state observer is used. and As a signal for transmission.

[0122] Design a reference signal estimator to obtain continuous state information:

[0123] (14)

[0124] in, and For position and velocity information estimated by the observer in the unknown state, It is a positive constant in the design and must satisfy the condition. ,in , This represents the minimum value of the set. for The symbolic function vector, similarly, for The symbolic function vector, and It can be represented as:

[0125] (15)

[0126] in, and Is the i-th drone at the trigger time The received position and velocity information of the j-th UAV; and Represents the i-th drone at the trigger time Received information on the location and speed of the leader drone.

[0127] (d) Establish an unknown state observer and estimate the unknown state after a deception attack.

[0128] Due to the existence of deception attacks, the drone's own state information is unreliable. Therefore, the following unknown state observer is constructed to estimate the true state information:

[0129] (16)

[0130] in, and This is an estimate of the actual position and velocity of the drone. and To estimate information for deceiving attack signals, It is an unknown item The estimated information, and For the designed diagonal positive matrix, The positive parameters for the design are m=1, 2, n=1, 2, 3. The optimal weight matrix for the neural network The estimated matrix, For Gaussian function vectors, The input signal is the Gaussian function vector, which the neural network uses to learn the function in (10). .

[0131] Learning using neural networks (10) , can be represented as:

[0132] (17)

[0133] in, for A real matrix of dimension , representing the optimal weight matrix. for A dimensional Gaussian function vector, For the input vector, To achieve the optimal learning error, This represents the number of nodes in the neural network. Represented as:

[0134] (18)

[0135] in, Represents the exponent Power of 1 For the center value vector, For width.

[0136] The estimated value in (16) is updated using an adaptive law, specifically as follows:

[0137] (19)

[0138] in, It is a positive constant for the design.

[0139] (e) Design a fractional-order sliding mode control method and construct a resilience control law.

[0140] Set the tracking error of the UAV state estimation signal to the estimated reference signal as follows: Based on this tracking error, the fractional-order sliding surface is designed as follows:

[0141] (20)

[0142] in, For constructing fractional-order sliding surfaces, For positive odd numbers, For a positive constant of design, for fractional derivative, It is a fractional operator. The definition of is:

[0143] (twenty one)

[0144] in, For fractional operators, express The nth derivative, It is the Gamma function.

[0145] Based on the fractional-order sliding surface, the following toughness control law is designed:

[0146] (twenty two)

[0147] in, and For the design of the positive parameter matrix, For the positive parameters of the design, m=1, 2, n=1, 2, 3; , for The absolute value of j, j=1, 2, 3; for The symbolic function vector, The integral symbol is used. express The second derivative vector.

[0148] To verify the effectiveness of this invention, the following experiments and analyses were conducted:

[0149] Design a swarm drone system consisting of four follower drones and one leader drone, with a topology that accommodates drones suffering from actuator failures and spoofing attacks, such as... Figure 1 As shown. The leader drone's flight path is set as follows: m, the distance between the follower drone and the leader drone is set to: m, m, m, m. Figure 2 The control framework for swarm drones was demonstrated.

[0150] The parameters for the swarm drones are set as follows: , , , , , , , . , , , , , , , , , , , , , , , , , , , , , , , , , , , .

[0151] The actuator failure suffered by the swarm drones is set as follows: Drone #1 in S suffers from actuator failure: Drone #1 in s suffers actuator failure Drone #3 in s suffers actuator failure Drone #4 in s suffers actuator failure The deception attack signals suffered by the drone are as follows:

[0152]

[0153]

[0154]

[0155]

[0156] Figure 3 The flight trajectories of the swarm drones are shown in the figure. It can be seen from the figure that the swarm drones can fly in good coordination. Although the flight trajectories of drone #1 and drone #2 deviated to some extent during the deception attack, the deviation quickly disappeared under the action of the control law and did not disrupt the coordinated formation of the swarm drones. Figure 4 The figure shows the location information of the swarm drones. As can be seen from the figure, the position change curves of the follower drones and the leader drones are basically the same, which proves the effectiveness of the designed resilience control scheme. Figure 5 The cooperative deviation signal of the follower drone is shown. It can be seen that when the drone is subjected to a deception attack, the cooperative deviation of the drone will fluctuate to a certain extent, but it remains within a certain range. After the deception attack ends, these cooperative deviations will disappear quickly, which fully demonstrates the effectiveness of the control scheme designed in this invention. Figure 6 The estimation bias of the unknown state observer designed in this invention is shown, and it can be seen that all tracking biases stabilize quickly. Although the estimation bias fluctuates somewhat during spoofing attacks, it remains within certain limits. Due to the event triggering mechanism, there is a small steady-state error in the z-axis direction, but this error has little impact on the cluster's coordination. Figure 7 The tracking deviation signal is shown. This tracking deviation is the tracking deviation in sliding mode control. As can be seen from the figure, the designed sliding mode control can achieve tracking of the estimated reference signal. Figure 8 As the actual control input signal is shown, after the deception attack ends, there will be a sudden change in the actual control input of the drone, but this change will disappear quickly, ensuring the stable flight of the drone. Figure 9 A comparison of the number of communication operations using the event-triggered mechanism and the time-triggered mechanism shows that the event-triggered communication mechanism requires only 20% to 25% of the number of communications compared to the time-triggered mechanism. In conclusion, when swarm drones with limited communication bandwidth suffer actuator failures and spoofing attacks, the observer-based event-triggered fractional-order adaptive resilient cooperative control method for swarm drones designed in this invention can ensure the safe cooperative flight of the swarm drones.

Claims

1. A clustered unmanned aerial vehicle (UAV) event-triggered fractional-order adaptive resilience cooperative control method, characterized in that, The method includes the following steps: Step 1: Establish the UAV dynamics model; Step 2: Based on the UAV dynamics model established in Step 1, construct the UAV dynamics model and actuator failure model under deception attack. Step 3: Design an event-triggered communication mechanism, establish a reference signal estimator, and estimate the UAV reference command signal; Step 4: Establish an unknown state observer to estimate the unknown state after a deception attack, and combine adaptive laws and neural network methods to learn the unknown terms caused by actuator failure. Step 5: Based on the reference command estimation information obtained in Step 3 and the unknown state estimation information obtained in Step 4, design a fractional sliding mode control method, construct a resilient control law, and realize the tracking of the reference signal.

2. The method according to claim 1, characterized in that, The UAV dynamics model established in step one is as follows: Set up a swarm of drones consisting of N followers and 1 leader, where the i-th drone has the following dynamics: (1) in, , and Let be the three-dimensional coordinates of the i-th UAV. express First derivative, express First derivative, express First derivative; For the speed of the drone, Represents the heading angle. Representing the flight path angle, the derivative is expressed as: (2) in, , , and The thrust, drag, lift, and lateral force of the drone are measured separately. It's an attack angle. It is the tilt angle. It is the sideslip angle; It's about the quality of the drone. It is the coefficient of gravitational acceleration; The thrust, drag, lift, and lateral force of a drone are expressed as follows: (3) in, It's the throttle opening. For maximum thrust, , It is air density. Wing area; , and These are the lift coefficient, drag coefficient, and lateral force coefficient, respectively, expressed by the following formulas: (4) in, , and It is a constant. , and To attack angle Correlation coefficient To match the sideslip angle Correlation coefficient; definition Input signal vector to the drone, Let the UAV's position vector be... For the drone's velocity vector, , , Let be the derivatives of the corresponding signs in the UAV position vector, respectively. Simplifying equations (1) and (2), we get: (5) in, and The transformation matrix is ​​represented as follows: (6)。 3. The method according to claim 2, characterized in that, Step two specifically involves: Considering the types of actuator failures faced by UAVs, the actuator failure model is defined as follows: (7) in, , It is the efficiency loss factor, j=1, 2, 3. This represents a deviation fault. The input signal deviation is caused by a fault, j=1, 2, 3. It is the UAV control input command signal; combined with equation (5), the UAV model under actuator failure is defined as: (8) in, The unknown item is caused by an actuator malfunction. It is the identity matrix; The drone is affected by a deception attack, resulting in unreliable position and velocity status information. The measurement signal of the drone after the deception attack is represented as follows: (9) in, and Unknown signals injected into deception attacks and Known position and velocity signals measured for the UAV; UAV position vector affected by deception attacks. and velocity vector It becomes an unknown signal; based on the UAV model (8) under actuator failure, combined with a deception attack, the following UAV model is obtained: (10) in, It is a known velocity signal measured by the drone. The known vector of influence, It is an unknown vector affected by a deception attack.

4. The method according to claim 3, characterized in that, Step three specifically involves: The cooperative bias of drones is defined as follows: (11) in, It is the communication weight between the i-th drone and the j-th drone. It is the communication weight between the i-th drone and the leader drone. This represents the distance between the i-th drone and the j-th drone. This represents the distance between the i-th drone and the leader drone. For the location information of the j-th UAV, Location information for the leader's drone; By modifying (11), the reference command signal for the i-th UAV is obtained. for: (12) Design an event-triggered communication mechanism: The i-th drone can only obtain the status information of its neighboring drones when an event is triggered; the first trigger time is set to... Design the following event to trigger the update condition for communication: (13) in, Representing the current moment, Indicates the trigger time The next triggering moment, Denotes the infimum of a function. For the design of positive constants, The velocity information is estimated by the unknown state observer; Due to deception attacks, the drone's own measured position and velocity signals... and speed signal Unreliable, using position and velocity information estimated by the observer in an unknown state. and As a transmitted signal; Design a reference signal estimator to obtain continuous state information: (14) in, and The reference position and velocity information estimated by the reference signal estimator. It is a positive constant in the design and must satisfy the condition. ,in , Represents the minimum value of the set; for The symbolic function vector, for The symbolic function vector, and Represented as: (15) in, and Is the i-th drone at the trigger time The received position and velocity information of the j-th UAV; and Represents the i-th drone at the trigger time Received information on the location and speed of the leader drone.

5. The method according to claim 4, characterized in that, The unknown state observer, adaptive law, and neural network established in step four are as follows: Construct the following unknown state observer to estimate the true state information: (16) in, and For position and velocity information estimated by the observer in the unknown state, and To estimate information for deceiving attack signals, It is an unknown item The estimated information, and For the designed diagonal positive matrix, For the positive parameters of the design, m=1, 2, n=1, 2, 3; The optimal weight matrix for the neural network The estimated matrix, For Gaussian function vectors, The input signal is the Gaussian function vector, which the neural network uses to learn the function in (10). .

6. The method according to claim 5, characterized in that, Learning using neural networks (10) , is represented as: (17) in, for A real matrix of dimension , representing the optimal weight matrix. for A dimensional Gaussian function vector, For the input vector, To achieve the optimal learning error, Represents the number of nodes in the neural network; Represented as: (18) in, Represents the exponent Power of 1 For the center value vector, Width; The estimated value in (16) is updated using an adaptive law, specifically as follows: (19) in, It is a positive constant for the design.

7. The method according to claim 6, characterized in that, The fractional sliding mode and toughness control laws in step five are as follows: Set the tracking error of the UAV state estimation signal to the estimated reference signal as: Based on this tracking error, the fractional-order sliding surface is designed as follows: (20) in, For constructing fractional-order sliding surfaces, For positive odd numbers, For a positive constant of design, for fractional derivative, It is a fractional operator; The definition of is: (21) in, For fractional operators, express The nth derivative, It is the Gamma function; Based on the fractional-order sliding surface, the following toughness control law is designed: (22) in, and For the design of the positive parameter matrix, For the positive parameters of the design, m=1, 2, n=1, 2, 3; , for The absolute value of j, j=1, 2, 3; for The symbolic function vector, The integral symbol is used. express The second derivative vector.

Images