An unmanned cluster consistency adaptive fault-tolerant control method for sensor and actuator mixed faults

Abstract

The application provides a kind of unmanned cluster consistency adaptive fault-tolerant control method under sensor and actuator mixed fault, wherein the method comprises: (1) converting the unmanned system model suffering from sensor and actuator mixed fault, establishing the augmented system in the form of pseudo actuator fault, so that the fault is reflected on the input channel; (2) constructing the formation coordination error model under the augmented system, and designing a system identifier based on neural network to approximate the nonlinear dynamics and lumped disturbance; (3) using the reinforcement learning algorithm based on Actor-Critic architecture, designing the Critic network update law with predefined time convergence property, and designing the optimal control term accordingly; (4) designing the fault compensation control term with predefined time based on the lumped disturbance approximation, combining it with the optimal control term, and introducing the Nussbaum function to construct the final consistency adaptive fault-tolerant controller. The application converts the sensor fault into pseudo actuator fault by model conversion method, reduces the compensation difficulty of sensor fault, effectively handles the model dynamics uncertainty by using reinforcement learning, and solves the problem of unknown sign of actuator fault coefficient by means of Nussbaum function, ensuring that the formation error of unmanned cluster can converge to a bounded region within a predefined time.

Description

Technical Field

[0001] This invention belongs to the field of collaborative control technology for multi-unmanned systems. Specifically, it relates to an adaptive fault-tolerant control method for unmanned cluster consistency under mixed sensor and actuator failures. Background Technology

[0002] With the continuous maturation of control technologies for multi-agent systems and unmanned systems, unmanned swarm systems, composed of multiple drones, unmanned vehicles, or unmanned ships, have been widely applied in engineering projects such as collaborative reconnaissance, material transportation, and area patrol due to their advantages of high efficiency, high flexibility, and the ability to replace humans in performing high-risk tasks. Among the key technologies of unmanned swarm systems, cooperative consistency control is the foundation for achieving overall behavior planning and complex tasks. Its core objective is to design a distributed control protocol so that the agents in the swarm can reach consistency in position, velocity, or other states by relying only on information exchange with their local neighbors. When the unmanned system malfunctions, traditional consistency control protocols designed based on ideal hardware conditions often fail, not only failing to maintain the cooperative performance of the swarm but also causing the entire system to diverge and collide. Therefore, researching fault-tolerant control methods for unmanned swarms has significant theoretical value and practical engineering implications.

[0003] In complex real-world operating environments, unmanned systems often suffer from physical losses at the actuator level, and sensor failures in the perception module are extremely common. Furthermore, the real-time convergence and anti-compound interference capabilities of the system are crucial to overall safety. For example, [Jing G, Liu C, Lan J, et al. Hierarchical Reinforcement Learning-BasedDistributed Fault-Tolerant Control for 2-D Plane Vehicular Platoon Systems[J]. IEEE Transactions on Vehicular Technology, 2025, 74(11): 16827-16838.], although it achieves fault-tolerant control for unknown direction failures of actuators in two-dimensional planar vehicle formations by combining a hierarchical reinforcement learning framework, its theoretical mechanism can only guarantee the semi-global consistency of the system error and eventual boundedness, failing to achieve the convergence effect within the preset time. Moreover, this scheme does not consider the problem of concurrent sensor and actuator failures under real-world conditions. This limitation means that when the cluster's sensing module and execution mechanism are damaged simultaneously, causing a compound fault, or when the system faces a risk avoidance scenario that requires a rapid response, the original fault tolerance mechanism not only cannot effectively handle the abnormal state feedback, but also fails due to a lack of effective convergence capability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a consensus-based adaptive fault-tolerant control method for unmanned swarms under mixed sensor and actuator faults. First, for unmanned systems with mixed faults, a low-pass filter is introduced to transform the model into an augmented system model, enabling fault tolerance for mixed sensor and actuator faults to be achieved simply by designing the controller on the input channel. A swarm cooperative error model is constructed based on the swarm formation, and a neural network is used to approximate the nonlinear dynamics and lumped disturbances in the system, designing corresponding system identifiers and state observers. Then, a reinforcement learning algorithm based on the Actor-Critic architecture is used to design a Critic network parameter update law with predefined time convergence properties to approximate the optimal cost function, thereby outputting an approximate optimal control action through the Actor network. Finally, a predefined time fault compensation control term is designed based on the lumped disturbance approximation, combined with the optimal control term, and a Nussbaum function is introduced to handle the sign uncertainty of actuator fault coefficients, constructing a complete consensus-based adaptive fault-tolerant controller.

[0005] To achieve the above objectives, this invention provides an adaptive fault-tolerant control method for unmanned swarm consensus under mixed sensor and actuator failures, comprising the following steps:

[0006] S1: Transform the unmanned system model suffering from sensor and actuator failures into an augmented system with pseudo-actuator failure characteristics;

[0007] S2: Constructing formation cooperative error under augmented systems and designing a system identifier based on neural networks;

[0008] S3: Design the optimal control terms and predefined time-based Critic network update laws based on the Actor-Critic architecture;

[0009] S4: Design a predefined time fault compensation control term and combine it with the designed Nussbaum function to obtain a consistent adaptive fault-tolerant controller.

[0010] Furthermore, the specific steps in step S1 of "converting the unmanned system model suffering from sensor and actuator failures into an augmented system with pseudo-actuator failure characteristics" are as follows:

[0011] S1a: Consider a containing A cluster of unmanned systems, where each unmanned system i contains a mixture of sensor and actuator faults, has the following system equations:

[0012]

[0013] in, , These represent state variables, inputs with actuator fault information, and system outputs, respectively. It is a perturbation that includes nonlinear terms. It is a bounded sensor fault message. These are the nonlinear terms of the system, A, B, C, D, E, It is a constant matrix, and the actuator fault is represented as: , These are bounded time-varying multiplicative coefficients, which can be positive or negative. It is a bounded time-varying deviation fault;

[0014] S1b: Introduce the following low-pass filter:

[0015] ,

[0016] in, The system equations in the augmented form are obtained as follows:

[0017]

[0018] in, , , , , , , , Through the above transformation, the sensor fault appears in the input channel of the system equation in the form of a nominal actuator fault. Through the content of the above step S1, the processing dimension of mixed faults is effectively unified, and the design difficulty of subsequent control laws is greatly reduced.

[0019] Furthermore, the specific steps in step S2, "constructing the formation cooperative error under the augmented system and designing a system identifier based on a neural network," are as follows:

[0020] S2a: Consider a... An unmanned swarm consisting of several unmanned systems tracks the virtual leader 0 in formation. The virtual leader's state is... The system equation for the virtual leader is:

[0021]

[0022] in The system matrix for the virtual leader is defined, along with the filtering state of the virtual leader. satisfy:

[0023] ,

[0024] The augmented virtual leader status is: ;

[0025] S2b: Based on existing undirected graph theory, define the formation coordination error between unmanned system i and its neighbor j, as well as between the virtual leader and other unmanned systems:

[0026] ,

[0027] in, Let be the element in the i-th row and j-th column of the adjacency matrix. For the i-th element in the virtual leader communication matrix, ,and , This represents the deviation of the unmanned system i from the ideal formation relative to the virtual leader. ,express The ideal formation spacing between unmanned systems i and j; let The derivative of the formation coordination error is obtained:

[0028]

[0029] S2c: Substituting the derivative of the formation coordination error and The expression yields the derivative of the formation coordination error that indicates the fault:

[0030]

[0031] S2d: Let the function for:

[0032] ,

[0033] in ;make The derivative of the reconstructed formation coordination error is:

[0034]

[0035] S2e: Approximated using a radial basis function neural network. Approaching The system identifier is obtained as follows:

[0036] ,

[0037] in It is the estimated value of the system identifier weight matrix. These are the basis functions of the system identifier;

[0038] S2f: Based on Design regarding auxiliary states State observer:

[0039]

[0040] in It is the observer gain. Yes The estimate, This is the adaptive fault-tolerant control law that needs to be designed later. Is for Robust compensation term; and define identification error as The design can prevent Divergent update law that maintains boundedness:

[0041] ,

[0042] in, Both are gain parameters. It is an exponential gain;

[0043] design The update law is:

[0044] ,

[0045] in For learning rate, Correction terms to prevent parameter drift;

[0046] design The expression is:

[0047]

[0048] in For a given minimum constant, and The above step S2 eliminates the reliance on a precise physical model of the unmanned system and enables real-time online identification of complex nonlinear dynamics and lumped disturbances caused by sensor failures.

[0049] Furthermore, the specific steps in step S3, "designing the optimal control term and predefined time-based Critic network update law based on the Actor-Critic architecture," are as follows:

[0050] S3a: Using a Critic network to approximate the optimal cost function :

[0051] ,

[0052] in For Critic network weights, These are the basis functions of the Critic network, and their gradients are defined. To minimize the formation coordination error and ensure the Critic network converges within a predefined time, the update law for the weights of the Critic network within a predefined time is designed as follows:

[0053] ,

[0054] in, and , and For setting Maximum value The convergence time is defined as follows. , And they are both even and odd numbers, , , , ;

[0055] S3b: According to optimal control theory, the ideal optimal control action is determined by the gradient of the Critic.

[0056] ;

[0057] Design the Actor network to output an approximate optimal control action:

[0058] ,

[0059] in For Actor network weights, These are the basis functions of the Actor network;

[0060] S3c: Design Actor Error :

[0061] ,

[0062] To minimize the error, the adaptive update law for Actor weights is:

[0063] ,

[0064] in For Actor learning rate, The content of step S3 above enables the adaptive weights of the Actor-Critic network to adjust within a set time. The network converges within a set time limit. This scheme introduces a predefined time-stabilized term based on fractional powers into the update law, enabling the network weights to approximate the network within a set time limit, which can effectively improve the rapid learning and response capabilities of unmanned systems in the face of sudden failures.

[0065] Furthermore, the specific steps in step S4, "designing a predefined time fault compensation control term and combining it with the designed Nussbaum function to obtain a consistent adaptive fault-tolerant controller," are as follows:

[0066] S4a: Design a fault compensation control term with a predefined time sliding surface and predefined time convergence properties:

[0067]

[0068] in for The pseudo-inverse matrix, The convergence time is defined as follows. The predefined time sliding surface is , The convergence time is defined as follows. , ;

[0069] S4b: Optimal control action and fault compensation control items Combining these, we obtain the complete control law:

[0070] ;

[0071] S4c: With actuator fault Time-varying multiplicative coefficients in the expression The error exhibits sign uncertainty, sometimes positive and sometimes negative. To address this sign uncertainty, a Nussbaum function is designed to ensure that the formation coordination error still converges to a bounded region near zero under these conditions. The specific form of the Nussbaum function is as follows:

[0072] ,

[0073] in, It is a natural constant. For the set constant, For Nussbaum adaptive parameters;

[0074] S4d: Designing Nussbaum Adaptive Parameters The update law is:

[0075] ;

[0076] S4e: Based on the Nussbaum function in step S4c, design a consistency-adaptive fault-tolerant controller as follows:

[0077] ,

[0078] in The designed consistent adaptive fault-tolerant controller Substituting this into the system equations of unmanned system i, we can ensure that, in the presence of mixed sensor and actuator failures, the formation coordination error of unmanned system i converges to a bounded region within a predefined time.

[0079] This invention addresses the challenges in current unmanned swarm cooperative control systems, such as the difficulty in simultaneously handling mixed sensor and actuator faults and poor real-time convergence. It provides a consistent adaptive fault-tolerant control method that enables unified handling of mixed faults and guarantees system convergence within a predefined timeframe. This novel control method not only reduces the difficulty of compensating for concurrent sensor and actuator faults but also achieves high-precision predefined-time formation tracking control even when the actuator fault coefficients have unknown signs and unknown nonlinear dynamic disturbances exist.

[0080] Compared with the prior art, the advantages of the present invention are as follows:

[0081] 1. This invention proposes a model transformation architecture. By introducing a low-pass filter, the unmanned system model suffering from mixed sensor and actuator faults is transformed into an augmented system in the form of pseudo-actuator faults. In this way, sensor faults appear in the input channel of the system equations as nominal actuator faults. Fault tolerance for mixed faults can be achieved simply by designing a controller on the input channel, which effectively reduces the difficulty of compensating for concurrent sensor faults under complex working conditions.

[0082] 2. In the face of nonlinear dynamic uncertainties and lumped disturbances in unmanned systems, this invention constructs a formation cooperative error model under augmented systems and designs a system identifier based on neural networks, effectively handling the model dynamic uncertainties; and based on the Actor-Critic reinforcement learning architecture, it designs a Critic network update law and optimal control term with predefined time convergence properties, realizing predefined time optimal adaptive control under unknown disturbances.

[0083] 3. Based on the lumped disturbance approximation and optimal control term, this invention constructs a fault compensation control term with a predefined time and combines it with the designed Nussbaum function, so that stable fault-tolerant control can still be performed even when the time-varying multiplicative coefficients with actuator faults have uncertainty in positive and negative signs. This ensures that the formation coordination error of the unmanned swarm can converge to a bounded region near zero within a predefined time when mixed faults exist. Attached Figure Description

[0084] Figure 1 This is a schematic diagram of the overall steps of the present invention.

[0085] Figure 2This is a schematic diagram of the formation tracking trajectory of the three followers in this embodiment.

[0086] Figure 3 This is a graph showing the formation coordination error norm of the three followers in this embodiment.

[0087] Figure 4 This is the norm curve of the estimated weight matrix of the system identifier for the three followers in this embodiment.

[0088] Figure 5 The curves represent the Critic network weight norms for the three followers in this embodiment.

[0089] Figure 6 The figure shows the Actor network weight norm curves for the three followers in this embodiment.

[0090] Figure 7 The norm curves of the fault compensation control terms for the three followers in this embodiment are shown. Detailed Implementation

[0091] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0092] Example: By Figure 1 As shown, the present invention provides an adaptive fault-tolerant control method for unmanned cluster consistency under mixed sensor and actuator failures. The specific implementation steps are as follows:

[0093] S1: Transform the unmanned system model suffering from sensor and actuator failures into an augmented system with pseudo-actuator failure characteristics;

[0094] S1a: Consider a containing A cluster of unmanned systems, where each unmanned system i contains a mixture of sensor and actuator faults, has the following system equations:

[0095]

[0096] in, , These represent state variables, inputs with actuator fault information, and system outputs, respectively. It is a perturbation that includes nonlinear terms. It is a bounded sensor fault message. These are the nonlinear terms of the system, A, B, C, D, E, It is a constant matrix, and the actuator fault is represented as: , These are bounded time-varying multiplicative coefficients, which can be positive or negative. It is a bounded time-varying deviation fault;

[0097] S1b: Introduce the following low-pass filter:

[0098] ,

[0099] in, The system equations in the augmented form are obtained as follows:

[0100]

[0101] in, , , , , , , , Through the above transformation, the sensor fault appears in the input channel of the system equation in the form of a nominal actuator fault.

[0102] S2: Constructing formation cooperative error under augmented systems and designing a system identifier based on neural networks;

[0103] S2a: Consider a... An unmanned swarm consisting of several unmanned systems tracks the virtual leader 0 in formation. The virtual leader's state is... The system equation for the virtual leader is:

[0104]

[0105] in The system matrix for the virtual leader is defined, along with the filtering state of the virtual leader. satisfy:

[0106] ,

[0107] The augmented virtual leader status is: ;

[0108] S2b: Based on existing undirected graph theory, define the formation coordination error between unmanned system i and its neighbor j, as well as between the virtual leader and other unmanned systems:

[0109] ,

[0110] in, Let be the element in the i-th row and j-th column of the adjacency matrix. For the i-th element in the virtual leader communication matrix, ,and , This represents the deviation of the unmanned system i from the ideal formation relative to the virtual leader. ,express The ideal formation spacing between unmanned systems i and j; let The derivative of the formation coordination error is obtained:

[0111]

[0112] S2c: Substituting the derivative of the formation coordination error and The expression yields the derivative of the formation coordination error that indicates the fault:

[0113]

[0114] S2d: Let the function for:

[0115] ,

[0116] in ;make The derivative of the reconstructed formation coordination error is:

[0117]

[0118] S2e: Approximated using a radial basis function neural network. Approaching The system identifier is obtained as follows:

[0119] ,

[0120] in It is the estimated value of the system identifier weight matrix. These are the basis functions of the system identifier;

[0121] S2f: Based on Design regarding auxiliary states State observer:

[0122]

[0123] in It is the observer gain. Yes The estimate, This is the adaptive fault-tolerant control law that needs to be designed later. Is for Robust compensation term; and define identification error as The design can prevent Divergent update law that maintains boundedness:

[0124] ,

[0125] in, Both are gain parameters. It is an exponential gain;

[0126] design The update law is:

[0127] ,

[0128] in For learning rate, Correction terms to prevent parameter drift;

[0129] design The expression is:

[0130]

[0131] in For a given minimum constant, and ;

[0132] S3: Design the optimal control terms and predefined time-based Critic network update laws based on the Actor-Critic architecture;

[0133] S3a: Using a Critic network to approximate the optimal cost function :

[0134] ,

[0135] in For Critic network weights, These are the basis functions of the Critic network, and their gradients are defined. To minimize the formation coordination error and ensure the Critic network converges within a predefined time, the update law for the weights of the Critic network within a predefined time is designed as follows:

[0136] ,

[0137] in, and , and For setting Maximum value The convergence time is defined as follows. , And they are both even and odd numbers, , , , ;

[0138] S3b: According to optimal control theory, the ideal optimal control action is determined by the gradient of the Critic.

[0139] ;

[0140] Design the Actor network to output an approximate optimal control action:

[0141] ,

[0142] in For Actor network weights, These are the basis functions of the Actor network;

[0143] S3c: Design Actor Error :

[0144] ,

[0145] To minimize the error, the adaptive update law for Actor weights is:

[0146] ,

[0147] in For Actor learning rate, ;

[0148] S4: Design a predefined time fault compensation control term and combine it with the designed Nussbaum function to obtain a consistent adaptive fault-tolerant controller;

[0149] S4a: Design a fault compensation control term with a predefined time sliding surface and predefined time convergence properties:

[0150]

[0151] in for The pseudo-inverse matrix, The convergence time is defined as follows. The predefined time sliding surface is , The convergence time is defined as follows. , ;

[0152] S4b: Optimal control action and fault compensation control items Combining these, we obtain the complete control law:

[0153] ;

[0154] S4c: With actuator fault Time-varying multiplicative coefficients in the expression The error exhibits sign uncertainty, sometimes positive and sometimes negative. To address this sign uncertainty, a Nussbaum function is designed to ensure that the formation coordination error still converges to a bounded region near zero under these conditions. The specific form of the Nussbaum function is as follows:

[0155] ,

[0156] in, It is a natural constant. For the set constant, For Nussbaum adaptive parameters;

[0157] S4d: Designing Nussbaum Adaptive Parameters The update law is:

[0158] ;

[0159] S4e: Based on the Nussbaum function in step S4c, design a consistency-adaptive fault-tolerant controller as follows:

[0160] ,

[0161] in The designed consistent adaptive fault-tolerant controller Substituting this into the system equations of unmanned system i, we can ensure that, in the presence of mixed sensor and actuator failures, the formation coordination error of unmanned system i converges to a bounded region within a predefined time.

[0162] To verify the effectiveness and effect of the method proposed in this invention, this embodiment uses Python for simulation verification, as detailed below:

[0163] In this implementation, the units for length and position information are set to meters (m), the unit for speed is meters per second (m / s), and the unit for time is seconds (s). N = 3 unmanned systems are used as followers in a formation to track the virtual leader's movement trajectory. The adjacency matrix between the followers is as follows:

[0164] ,

[0165] The virtual leader communication matrix is , , , The initial positions of the followers are respectively , , This involves setting up three followers in a triangular formation to track the virtual leader.

[0166] In the simulation verification here, to facilitate the demonstration of the good control effect of the method proposed in this invention, the following settings are provided. , , , , , The actuator failure modes for followers 1-3 are set as follows:

[0167]

[0168]

[0169]

[0170] The sensor failure modes for followers 1-3 are set as follows:

[0171]

[0172]

[0173]

[0174] The system matrix for setting the virtual leader is as follows:

[0175] ,

[0176] in The value is rad / s and is constant. Other variable values ​​are set as follows: , , , , , , , , , , , , , , , , , , , , , , , , The basis functions of the system identifier, Actor network, and Critic network all adopt two-dimensional Gaussian fuzzy basis functions, and the simulation time is set to 60s.

[0177] Figure 2 The diagram shows the formation tracking trajectory of three followers. Figure 2 It is evident that the three followers can maintain their formation well even when a malfunction occurs, and their trajectories will not diverge. They can maintain the predetermined triangular formation until the end of the mission.

[0178] Figure 3 A graph showing the norm of the formation coordination error for the three followers. Figure 3 It is evident that the three followers can maintain formation coordination error without continuous divergence even in the event of a fault, and converge to near zero within 5 seconds of the fault change.

[0179] Figure 4 , Figure 5 , Figure 6 These figures represent the norm curves of the weight matrix estimates for the three followers' collective identifiers, the norm curves of the Critic network's weights, and the norm curves of the Actor network's weights. As can be seen from the three figures, all weight values ​​remain bounded and do not diverge when different faults occur, and convergence is achieved once the fault no longer changes. In particular, the norm curve of the Critic network's weights converges within a predetermined 3 seconds.

[0180] Figure 7 The figure shows the norm curves of the fault compensation control terms for the three followers. As can be seen from the figure, each fault change causes a sharp rise in the curve, thereby achieving rapid fault compensation and thus realizing stable formation tracking control.

[0181] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A method for adaptive fault-tolerant control of unmanned swarm consensus under mixed sensor and actuator faults, characterized in that... This includes the following steps: S1: Transform the unmanned system model suffering from sensor and actuator failures into an augmented system with pseudo-actuator failure characteristics; S2: Constructing formation cooperative error under augmented systems and designing a system identifier based on neural networks; S3: Design the optimal control terms and predefined time-based Critic network update laws based on the Actor-Critic architecture; S4: Design a predefined time fault compensation control term and combine it with the designed Nussbaum function to obtain a consistent adaptive fault-tolerant controller.

2. The adaptive fault-tolerant control method for unmanned swarm consistency under mixed sensor and actuator failures as described in claim 1, characterized in that, The specific steps of step S1 are as follows: S1a: Consider a containing A cluster of unmanned systems, where each unmanned system i contains a mixture of sensor and actuator faults, has the following system equations: in, , These represent state variables, inputs with actuator fault information, and system outputs, respectively. It is a perturbation that includes nonlinear terms. It is a bounded sensor fault message. These are the nonlinear terms of the system, A, B, C, D, E, It is a constant matrix, and the actuator fault is represented as: , These are bounded time-varying multiplicative coefficients, which can be positive or negative. It is a bounded time-varying deviation fault; S1b: Introduce the following low-pass filter: ， in, The system equations in the augmented form are obtained as follows: in, , , , , , , , Through the above transformation, the sensor fault appears as a nominal actuator fault on the input channel of the system equation.

3. The adaptive fault-tolerant control method for unmanned swarm consensus under mixed sensor and actuator failures as described in claim 1, characterized in that... The specific steps of step S2 are as follows: S2a: Consider a... An unmanned swarm consisting of several unmanned systems tracks the virtual leader 0 in formation. The virtual leader's state is... The system equation for the virtual leader is: in The system matrix for the virtual leader is defined, along with the filtering state of the virtual leader. satisfy: ， The augmented virtual leader status is: ; S2b: Based on existing undirected graph theory, define the formation coordination error between unmanned system i and its neighbor j, as well as between the virtual leader and other unmanned systems: , in, Let be the element in the i-th row and j-th column of the adjacency matrix. For the i-th element in the virtual leader communication matrix, ,and , This represents the deviation of the unmanned system i from the ideal formation relative to the virtual leader. ,express The ideal formation spacing between unmanned systems i and j; let The derivative of the formation coordination error is obtained: S2c: Substituting the derivative of the formation coordination error and The expression yields the derivative of the formation coordination error that indicates the fault: S2d: Let the function for: in ;make The derivative of the reconstructed formation coordination error is: S2e: Approximated using a radial basis function neural network. Approaching The system identifier is obtained as follows: ， in It is the estimated value of the system identifier weight matrix. These are the basis functions of the system identifier; S2f: Based on Design regarding auxiliary states State observer: in It is the observer gain. Yes The estimate, This is the adaptive fault-tolerant control law that needs to be designed later. Is for Robust compensation term; and define identification error as The design can prevent Divergent update law that maintains boundedness: , in, Both are gain parameters. It is an exponential gain; design The update law is: ， in For learning rate, Correction terms to prevent parameter drift; design The expression is: in For a given minimum constant, and .

4. The adaptive fault-tolerant control method for unmanned swarm consistency under mixed sensor and actuator failures as described in claim 1, characterized in that, The specific steps of step S3 are as follows: S3a: Using a Critic network to approximate the optimal cost function : ， in For Critic network weights, These are the basis functions of the Critic network, and their gradients are defined. To minimize the formation coordination error and ensure the Critic network converges within a predefined time, the update law for the weights of the Critic network within a predefined time is designed as follows: ， in, and , and For setting Maximum value The convergence time is defined as follows. , And they are both even and odd numbers, , , , ; S3b: According to optimal control theory, the ideal optimal control action is determined by the gradient of the Critic. ; Design the Actor network to output an approximate optimal control action: ， in For Actor network weights, These are the basis functions of the Actor network; S3c: Design Actor Error : ， To minimize the error, the adaptive update law for the Actor network weights is: ， in For Actor learning rate, .

5. The adaptive fault-tolerant control method for unmanned swarm consensus under mixed sensor and actuator failures as described in claim 1, characterized in that, The specific steps of step S4 are as follows: S4a: Design a fault compensation control term with a predefined time sliding surface and predefined time convergence properties: in for The pseudo-inverse matrix, The convergence time is defined as follows. The predefined time sliding surface is , The convergence time is defined as follows. , ; S4b: Optimal control action and fault compensation control items Combining these, we obtain the complete control law: ； S4c: With actuator fault Time-varying multiplicative coefficients in the expression The error exhibits sign uncertainty, sometimes positive and sometimes negative. To address this sign uncertainty, a Nussbaum function is designed to ensure that the formation coordination error still converges to a bounded region near zero under these conditions. The specific form of the Nussbaum function is as follows: , in, It is a natural constant. For the set constant, For Nussbaum adaptive parameters; S4d: Designing Nussbaum Adaptive Parameters The update law is: ； S4e: Based on the Nussbaum function in step S4c, design a consistency-adaptive fault-tolerant controller as follows: ， in The designed consistent adaptive fault-tolerant controller Substituting this into the system equations of unmanned system i, we can ensure that, in the presence of mixed sensor and actuator failures, the formation coordination error of unmanned system i converges to a bounded region within a predefined time.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements an adaptive fault-tolerant control method for unmanned cluster consistency under mixed sensor and actuator failures, as described in any one of claims 1 to 5.

7. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instruction is executed by the processor, it implements an adaptive fault-tolerant control method for unmanned cluster consistency under mixed sensor and actuator failures as described in any one of claims 1-5.

Images