A distributed fixed-wing unmanned aerial vehicle preset performance fault-tolerant formation control method

By adopting a distributed fixed-wing UAV pre-set performance fault-tolerant formation control method, the problems of insufficient shape maintenance and fault tolerance of UAV formations in complex environments are solved, and stable operation and efficient control of UAV formations under actuator failure and disturbance are achieved.

CN123067487APending Publication Date: 2026-11-10BEIHANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511388839.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-09-26
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing drone formation control methods struggle to maintain formation shape in complex or signal-constrained environments and lack effective fault-tolerance compensation mechanisms, failing to meet the safety and performance predictability requirements of engineering needs.

Method used

A distributed fixed-wing UAV pre-set performance fault-tolerant formation control method is adopted. By establishing a nonholonomic dynamic model, designing a time-defined observer and hierarchical error transformation, and combining local projection matrix and error feedback compensation, the UAV formation is ensured to operate stably under actuator failure and environmental disturbance.

Benefits of technology

It achieves stability and fault tolerance of UAV formations in resource-constrained and complex environmental scenarios, significantly improves the robustness and computational efficiency of the system, and ensures that the error trajectory is always within the preset performance boundaries, thus meeting engineering requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123067487A_ABST
    Figure CN123067487A_ABST
Patent Text Reader

Abstract

The application relates to the field of cooperative control and formation flight of unmanned aerial vehicles, and discloses a distributed fixed-wing unmanned aerial vehicle preset performance fault-tolerant formation control method, which comprises the following steps: S1. System modeling and problem description: a non-complete dynamic model of a fixed-wing unmanned aerial vehicle is established, and uncertain factors such as actuator faults, thrust saturation and environmental disturbances are considered and unified as cluster uncertainties; S2. Uncertainty observer design; S3. Hierarchical error transformation; S4. Fault-tolerant control law design based on position constraint. By only relying on the position information of adjacent unmanned aerial vehicles, the application avoids the dependence on a global reference coordinate and a high-precision positioning system, so that the control method can stably work in a task scene with limited resources and a complex environment. Meanwhile, the prescribed time observer effectively compensates for actuator faults, thrust saturation, environmental disturbances and parameter uncertainties, and significantly improves the fault tolerance and robustness of the system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of collaborative control and formation flight technology for unmanned aerial vehicles (UAVs), specifically a distributed fixed-wing UAV preset performance fault-tolerant formation control method. Background Technology

[0002] With the widespread application of UAV swarms in disaster relief, military reconnaissance, environmental monitoring, and logistics transportation, formation control technology has gradually become a core issue in UAV swarm research. Existing formation control methods mainly include control frameworks based on displacement information, distance information, and azimuth information. Displacement-based control methods typically require each UAV to obtain globally unified reference coordinates or absolute position, which usually relies on satellite navigation systems or high-precision positioning equipment. However, due to signal limitations, especially in complex or signal-constrained environments, the application of this method is greatly restricted. Distance-based control methods reduce reliance on positioning to some extent, but during formation translation or scaling maneuvers, distance constraints cannot maintain the invariance of the formation shape, easily leading to formation degradation. Therefore, distance-based control methods have limitations in complex maneuver applications.

[0003] In contrast, formation control methods based on azimuth information only require data interaction between the relative azimuth information of adjacent UAVs, without relying on global reference coordinates. This gives azimuth-based control methods a significant advantage in resource-constrained and complex mission scenarios, especially in environments where satellite navigation or high-precision positioning equipment cannot be used. Nevertheless, existing azimuth-based formation control methods still have significant shortcomings. On the one hand, many studies design control laws based on idealized first-order or linear models, ignoring the nonholonomic and underactuated dynamic characteristics of fixed-wing UAVs themselves. This makes it difficult to directly apply theoretical results to practical systems. On the other hand, most existing methods can only achieve asymptotic or finite-time convergence of errors, and cannot effectively constrain the transient and steady-state behavior of errors during the convergence process, thus failing to meet the safety and performance predictability requirements in engineering needs.

[0004] Furthermore, when the drone's actuators experience multiplicative or additive failures, or when thrust saturation, system parameter perturbations, or external environmental disturbances occur, traditional controllers often lack effective fault-tolerant compensation mechanisms, which may cause the drone to deviate from the expected trajectory or even disrupt the stability of the entire formation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a distributed fixed-wing unmanned aerial vehicle (UAV) preset performance fault-tolerant formation control method, which solves the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a distributed fixed-wing unmanned aerial vehicle (UAV) pre-set performance fault-tolerant formation control method, comprising the following steps:

[0007] S1. System Modeling and Problem Description: Establish a nonholonomic dynamic model of the fixed-wing UAV, considering uncertainties such as actuator failure, thrust saturation, and environmental disturbances, and unify them into cluster uncertainties;

[0008] S2. Uncertainty Observer Design: For the introduced cluster uncertainty, a time-defined observer is established, and a time scaling function is used to establish the update law for uncertainty estimation to ensure accurate estimation within the specified time.

[0009] S3. Hierarchical error transformation: Set the position error and set the position-velocity composite error on this basis. Apply control constraints to the position error and the position-velocity composite error. Control the position error trajectory through a monotonically decreasing performance function to ensure that the error is always within the preset performance boundary.

[0010] S4. Design of fault-tolerant control law based on orientation constraint: By using a local projection matrix, the error signal is projected into the orientation constraint subspace, thereby filtering out invalid control inputs. The control law is constructed in combination with the error, and the stability of the UAV formation is ensured through error feedback and feedforward compensation.

[0011] Preferred options also include:

[0012] S5. Stability Analysis and Parameter Condition Proof: The closed-loop stability of the system is rigorously proven using the Lyapunov method, and the conditions for the value of the control gain are given to ensure that the formation can still operate stably under external disturbances and actuator failures.

[0013] Preferably, the nonholonomic dynamics model of the fixed-wing UAV includes the kinematic and dynamic equations of each UAV, describing the UAV's flight path angle, heading angle, mass, drag, environmental disturbances, and control inputs.

[0014] Preferably, the cluster uncertainty sets characterize the combined effects of actuator multiplicative faults, additive biases, thrust saturation errors, parameter perturbations, and external disturbances.

[0015] Preferably, S1 specifically includes:

[0016] S101. Establish a dynamic model of a fixed-wing UAV. Each UAV is modeled as a rigid body, and its motion is described by both the inertial coordinate system and the body coordinate system. The kinematic and dynamic equations of the i-th UAV are as follows:

[0017]

[0018] Where, p i =[x i ,y i ,z i ] T V represents the position vector. i Indicates airspeed, γ i and ψ i These are the flight path angle and heading angle, respectively. For actual quality, D i As resistance, d ei For environmental interference, U i L i and μ i These are the control inputs for thrust, lift, and roll angle, respectively.

[0019] S102. Set the formation topology as leader-first follower, and the remaining followers establish perception with local neighbors through orientation information, and construct an orthogonal projection matrix using local orientation information.

[0020] Preferably, in the topology, the first follower only perceives the leader, while other followers perceive the neighbors with smaller numbers, and ensures that the orientation constraints are not collinear, and that all orientation constraints conform to the center-to-scale geometric description.

[0021] Preferably, the specified time observer ensures that the system error converges within a finite time by introducing a time scaling function and a dynamic gain term.

[0022] Preferably, the control law combines error feedback compensation terms with feedforward compensation terms to ensure the stability and fault tolerance of the formation when performing complex maneuvers, without relying on the global coordinate system, but only on local neighbor azimuth information.

[0023] Preferably, the update law of the time-defined observer is:

[0024]

[0025] Where, α 1i ,α 2i >0,β 1i ,β 2i >4 is the gain for the observer design.

[0026] This invention provides a distributed fixed-wing unmanned aerial vehicle (UAV) pre-set performance fault-tolerant formation control method. It has the following beneficial effects:

[0027] 1. By relying solely on the azimuth information of neighboring UAVs, this invention avoids dependence on global reference coordinates and high-precision positioning systems, enabling the control method to operate stably in resource-constrained and complex environments. Simultaneously, the use of a time-defined observer effectively compensates for actuator failures, thrust saturation, environmental disturbances, and parameter uncertainties, significantly improving the system's fault tolerance and robustness.

[0028] 2. This invention employs a hierarchical performance error transformation, which can strictly constrain the transient and steady-state behavior of position error and position-velocity composite error, ensuring that the error trajectory remains within the preset performance boundaries. This performance constraint ensures precise error control during the dynamic flight of UAV formations, improving system safety and mission reliability.

[0029] 3. The control law of this invention significantly reduces computational complexity by utilizing local orientation information. Furthermore, the control parameters are easy to verify, providing feasibility for large-scale UAV swarm applications. The distributed design of this method enables stable operation across multiple UAV swarms, demonstrating good scalability and broad engineering application prospects. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the process of the present invention;

[0031] Figure 2 This is a cube communication topology diagram of the present invention, which includes 8 drones;

[0032] Figure 3 This is a schematic diagram illustrating the control framework of the present invention;

[0033] Figure 4 This is a schematic diagram of the position tracking error transformation of the present invention;

[0034] Figure 5 This is a schematic diagram of the formation trajectory of 8 drones according to the present invention;

[0035] Figure 6 This is a schematic diagram of the position tracking error of the follower of the present invention changing over time;

[0036] Figure 7 This is a graph showing the change in the combined tracking error of the followers of the present invention over time.

[0037] Figure 8 This is a graph showing the control input of the present invention changing over time.

[0038] Figure 9 This is a graph showing the change of the concentrated uncertainty error of the present invention over time. Detailed Implementation

[0039] The technical solutions in 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.

[0040] Example:

[0041] See attached document Figures 1-9 A distributed fixed-wing UAV pre-set performance fault-tolerant formation control method includes the following steps:

[0042] S1. System Modeling and Problem Description: This section primarily establishes a nonholonomic dynamic model of the fixed-wing UAV, comprehensively considering uncertainties such as thrust saturation, multiplicative and additive actuator faults, and environmental disturbances. These uncertainties are uniformly modeled as clustered uncertainties, laying the theoretical foundation for subsequent observer design, error transformation, and fault-tolerant control laws. Specifically:

[0043] S101. Dynamic Modeling and Assumption Setting: A dynamic model of a fixed-wing UAV is established, considering basic parameters such as thrust, lift, mass, and drag. Each UAV is modeled as a rigid body, with its motion described by both inertial and body coordinate systems. Uncertainties in the system are assumed to include actuator failures (multiplicative and additive failures), thrust saturation, and environmental disturbances. Specifically: multiplicative actuator failures stem from physical degradation such as motor efficiency decline and propeller damage, manifesting as a scaling factor in thrust commands; additive failures are mostly caused by actuator circuit misalignment and mechanical jamming, manifesting as constant bias output; thrust saturation is the physical limit of the maximum thrust of the power system, such as the maximum power limit of the motor, manifesting as clamping characteristics of input commands; environmental disturbances mainly include atmospheric interference such as gusts and turbulence, exhibiting bounded characteristics and bounded rates of change. These factors are normalized into cluster uncertainties, providing a theoretical basis for subsequent steps.

[0044] A dynamic model of a fixed-wing UAV is established, considering a system consisting of n fixed-wing UAVs. Each UAV is modeled as a rigid body, and its motion is described by both an inertial coordinate system (describing absolute position and attitude) and a body coordinate system (characterizing thrust, lift, and other forces and moments). Taking the i-th UAV in the formation as an example, its kinematic and dynamic equations cover core states such as position vector, airspeed, flight path angle, and heading angle, while incorporating basic physical parameters such as mass, drag, and lift. The kinematic and dynamic equations of the i-th UAV are shown below:

[0045]

[0046] Where, p i =[x i ,yi ,z i ] T V represents the position vector. i Indicates airspeed, γ i and ψ i These are the flight path angle and heading angle, respectively. For actual quality, D i As resistance, d ei For environmental interference, U i L i and μ i The system configuration can be simplified by the following coordinate transformation steps, where thrust, lift, and roll angle are the control inputs respectively:

[0047] 1. Define the unit vector for the velocity direction:

[0048] In the body coordinate system, the direction of velocity can be determined by the path angle and the heading angle:

[0049]

[0050] This vector represents the direction the aircraft is facing.

[0051] 2. Position kinematics relationship:

[0052] From the kinematic relationship, we can obtain:

[0053] That is, the change in position is equal to the product of airspeed and direction vector.

[0054] 3. Directional dynamics relationship:

[0055] The change in direction vector is driven by angular velocity, angular velocity vector

[0056] Remember [ω] i ] × Let be the corresponding antisymmetric matrix, then we have:

[0057] 4. Airspeed dynamics relationship:

[0058] Airspeed is affected by thrust, drag, and disturbance, and satisfies:

[0059] Where τ i =U i cosμ i -D i This indicates the combined thrust / drag effect.

[0060] The system can ultimately be simplified to the following second-order nonholonomic form:

[0061]

[0062] in, τ represents the heading vector. i For longitudinal acceleration input, ω i This is the angular velocity vector. [ω] i ] × This indicates that the antisymmetric matrix operator satisfies [ω]. × x = ω × x.

[0063] S102. Topology and Orientation Information Modeling: The formation topology is set as leader-first follower. Other followers establish perception with their local neighbors through orientation information. An orthogonal projection matrix is ​​constructed using local orientation information. A "center-scale" geometric description is introduced, defining the desired formation center and scale function to ensure the stability of the formation during translation and contraction. Specifically, this includes:

[0064] The formation topology is set as a leader-first follower mode. The desired positions and speeds of the leader and first follower are directly set by the task, while the remaining followers perceive the positions through relative orientation information with their local neighbors, thereby achieving distributed formation control.

[0065] To achieve error measurement when only orientation information is available, let the actual relative orientation vector between the i-th UAV and its neighbor j be:

[0066]

[0067] Where, p i (t),p j (t) represents the positions of drone i and neighbor j, respectively. It represents a three-dimensional unit sphere.

[0068] And thus construct the orthogonal projection matrix:

[0069]

[0070] Specifically, by subtracting the outer product of the azimuth vector and its transpose from the third-order identity matrix, the orthogonal projection matrix can be obtained. This matrix can map any three-dimensional vector to a two-dimensional subspace orthogonal to the azimuth vector, thereby effectively filtering out unidentifiable components along the azimuth direction and retaining only the lateral components that constrain the formation shape.

[0071] After constructing the orthogonal projection matrix, this invention uses this matrix to orthogonally decompose the system error to clearly define signals such as position error, velocity error, and composite error. Specifically, the deviation of each UAV from its desired position is first defined as the position tracking error, i.e., the difference between the actual position vector and the desired position vector. Considering that formation control relying solely on azimuth information cannot directly observe radial deviations, to avoid the influence of invalid or unmeasurable error components on control, this invention projects the aforementioned position error onto a subspace orthogonal to the azimuth constraint using the orthogonal projection matrix. This orthogonal projection matrix filters out the components of the error vector in the neighboring azimuth direction, ensuring that the extracted error signal only contains the effective part perpendicular to the relative azimuth, thereby discarding part of the position error along the line-of-sight direction. Thus, the error decomposition focuses on the lateral deviation affecting the formation geometry, avoiding redundant control inputs in unobservable directions.

[0072] Based on the above error decomposition, this invention strictly limits the allowable range of each error signal and introduces preset performance constraints to control the error trajectory. For position error, an allowable upper bound is given at the initial moment, and a monotonically decreasing performance function is designed as a contraction boundary, so that the upper bound gradually contracts to the desired steady-state value over time. The position error must always remain within this dynamic boundary.

[0073] The constraint is: |e pi (t)|<ρ i (t), where ρ i (t) represents the performance function described above.

[0074] ρ can be selected i (t) is an exponential decay function, with its initial value being the allowable upper bound of the initial error, its steady-state value being the allowable residual error, and its decay rate set according to task requirements. To incorporate this inequality constraint into the control law, this invention further defines a nonlinear error transformation variable, mapping the position error to a confined space. By utilizing logarithmic and hyperbolic tangent functions for error transformation, it can be deduced that when the transformed error variable remains bounded, the original position error constantly satisfies...

[0075] In other words, the controller only needs to ensure the stable convergence of the transformation error to ensure that the position error remains within the preset performance boundaries throughout the entire process and will not exceed these boundaries during convergence. This preset performance constraint design guarantees that the entire trajectory of the position tracking error, from the initial transient state to the final steady state, remains within a safe range.

[0076] Building upon the constraints imposed on position errors, this invention further defines velocity error and composite error to improve the consistency and convergence characteristics of formation motion. Velocity error is defined as the difference between the actual velocity of the i-th UAV and the desired formation velocity, used to measure velocity tracking deviation. Furthermore, a position-velocity composite error is introduced as a weighted combination of position and velocity errors. This composite error comprehensively reflects spatial position deviation and velocity inconsistency, compensating for information lost due to azimuth control alone. Specifically, since the orthogonal projection matrix filters out radial distance errors, simple position errors cannot capture the impact of distance changes between UAVs. This invention introduces velocity error to reflect formation scaling motion, compensating for information loss caused by orthogonal projection. The composite error signal includes the projected position and velocity differences between adjacent UAVs, revealing dynamic inconsistencies within a local neighborhood. This invention also imposes contractionary performance boundary constraints on each component of the composite error, selecting a monotonically decreasing performance function and performing nonlinear error transformation to ensure that the composite error consistently meets preset transient and steady-state performance indicators throughout the entire flight mission. This hierarchical error transformation and constraint design extends the constraints from single position errors to joint position-velocity errors, enabling synchronized position convergence and velocity synchronization in UAV formations. This not only ensures gradual convergence of position errors but also enhances velocity synchronization, avoiding formation distortion and divergence caused by velocity differences in traditional methods.

[0077] like Figure 2 As shown, this implementation uses a leader-first follower structure. The leader is node 1, the first follower is node 2, and each of the remaining followers perceives exactly two neighbors with smaller topological numbers. It satisfies the following conditions: First, the first follower perceives only the leader; second, all other followers perceive only two neighbors with smaller topological numbers; third, all directional constraints are not collinear.

[0078] Considering the effects of actuator failure and input saturation, the actual model for longitudinal acceleration is as follows:

[0079] τ i,F =ρ i τ i,c +φ i

[0080] Where τ i,c For the virtual longitudinal acceleration command issued by the controller, ρ i It is a multiplicative fault factor, whose characteristics are unknown and positively bounded, ρ i The additive fault bias has unknown and bounded characteristics. The actual effect of τ on the aircraft is... i Subject to saturation mapping:

[0081] τ i=sat(τ i,F )

[0082] And define the saturation cutoff value.

[0083] Δ i =τ i,F -τ i

[0084] Meanwhile, model error, environmental disturbance, and other terms are uniformly denoted as cluster uncertainty d. i (t), which is bounded and has a bounded rate of change.

[0085] When establishing the dynamic model of the UAV formation system described in this invention, the uncertainty factor d is considered. i (t) makes the following assumptions:

[0086] 1. Factors causing actuator failure:

[0087] The thrust actuator of a drone may experience multiplicative and additive failures.

[0088] Multiplicative faults are represented by the proportional coefficient ρ i The characterization is defined as follows, and its value ranges from [0,1]. When ρ i =1 indicates that the actuator is completely normal, when ρ i =0 indicates that the thrust is completely ineffective.

[0089] Additive faults with constant bias φ i Characterization is used to describe the fixed difference between the actuator output and the desired instruction.

[0090] 2. Thrust saturation factor:

[0091] Considering the physical limits of UAV thrust, the thrust command is subject to saturation constraints. Assume the maximum thrust is U. max When the actual control input exceeds this limit, the actuator output is fixed at the boundary value. Thrust saturation will lead to a decrease in control accuracy and introduce additional dynamic errors.

[0092] 3. Parameter perturbation factors:

[0093] UAV mass parameter m i Perturbations may occur during flight due to fuel consumption or changes in payload. Let the actual mass be... Where Δ i This is the mass deviation. The uncertainty of this parameter directly affects the accuracy of acceleration calculation and manifests as modeling error in system dynamics.

[0094] 4. External disturbance factors:

[0095] Environmental disturbances are denoted as This includes wind field disturbances, airflow variations, and random noise. The disturbance is a time-varying bounded signal, with its upper amplitude denoted as... This disturbance is directly added to the drone's acceleration equation, affecting its trajectory tracking performance.

[0096] Cluster uncertainty d i (t) is defined as:

[0097]

[0098] The uncertainty d of the cluster i (t) centrally characterizes the actuator multiplicative fault ρ i Additive bias φ i Thrust saturation error Δ i The combined effects of factors such as parameter perturbations and external disturbances. Under the above assumptions, the defined cluster uncertainty d i (t) uniformly reflects the combined effects of various uncertainties such as actuator multiplicative faults, additive bias, thrust saturation, parameter perturbations, and external disturbances. This uncertainty is estimated and compensated for in the control design using a time-defined observer, thereby ensuring the consistency and stability of formation control under various uncertain conditions.

[0099] In this embodiment, the desired formation is modeled using a "center-scale" geometric description. Therefore, the center of the desired formation is first defined:

[0100]

[0101] in, Let n represent the expected position vector of the i-th UAV at time t, where n is the formation size. Simultaneously, a scaling function is introduced:

[0102]

[0103] Used to describe the expansion and contraction characteristics of the formation as a whole.

[0104] Within this framework, the expected velocity of the leader and the first follower is designed as follows:

[0105]

[0106] Among them, v c (t) is the translational velocity of the formation center, and α(t) is the scale change rate, which is used to realize the overall expansion and contraction motion of the formation.

[0107] The aforementioned velocity model ensures that the leader and first follower can simultaneously reflect translation and scale changes during trajectory planning, thereby providing a baseline constraint for the desired geometry of the entire formation.

[0108] For the desired positions of the remaining followers, the local location constraint set is used:

[0109]

[0110] The relative configuration relationship is uniquely determined. That is, the non-leader-first follower drone only needs to obtain the expected orientation vectors of its neighboring nodes to recover its own expected position, thereby avoiding dependence on the global coordinate system.

[0111] To ensure that the distributed formation maneuver guided by the leader and the first follower has a unique solution, a leader-first follower identifiability condition needs to be introduced. The core idea is that, given the desired positions and velocities of the leader and the first follower, the remaining followers can uniquely determine their positions within the desired formation through local location constraints.

[0112] The expected positions of the remaining followers do not depend on global coordinates and can be uniquely determined solely by the expected orientation vectors of their neighbors: the expected states of the leader and the first follower are known, each follower obtains at least two non-collinear orientation information, the topology remains connected and there are directed paths from the leader to all followers, fundamentally eliminating the dependence on a global navigation system and providing a geometric basis for distributed control.

[0113] Formalistically, let G = (V, ξ) be the communication-aware topology of the formation, where V = {1, 2, ..., n}. A formation is said to satisfy the leader-first follower identifiability condition if the following conditions are met:

[0114] 1. The expected state of the leader as node 1 and as the first follower as node 2. Known;

[0115] 2. For each of the remaining followers i≥3, there exist at least two edges (i,j1), (i,j2)∈ξ such that the corresponding desired orientation vector Not collinear;

[0116] 3. The topology graph G maintains connectivity and contains a directed path from the leader to each follower.

[0117] Under these conditions, it can be proven that for any time t, the expected position of the non-leader—the first follower—is... It can be uniquely determined through local orientation information, ensuring the identifiability of the desired formation geometry.

[0118] By employing a center-scale desired formation modeling approach, combined with leader-first-follower desired velocity design and leader-first-follower identifiability conditions, the translation and scale-joint maneuvers of the desired formation can be achieved without global reference coordinates. This design not only reduces dependence on external navigation systems but also ensures the uniqueness and recoverability of the formation geometry, thus laying the foundation for the design of distributed fault-tolerant control laws.

[0119] S2. Design of a Specified-Time Uncertainty Observer. A specified-time observer is designed to address cluster uncertainties introduced by actuator multiplicative and additive faults, thrust saturation cutoff, parameter uncertainties, and external disturbances. An update law for velocity and uncertainty estimation is established by introducing a time scaling function.

[0120] To ensure safe and reliable formation maneuvers under conditions of failure, bias actuator malfunction, and multiple system uncertainties, this invention proposes a cascaded control framework with specified performance, see appendix. Figure 3 Specifically, it includes:

[0121] S201. Definition of Error and Uncertainty: Define the rate and uncertainty estimation error; design a time-defined observer for accurate estimation of uncertainties in the system; introduce a time scaling function to accelerate the convergence of the error within a finite time by dynamically adjusting the gain term.

[0122] In this implementation method, various uncertainties existing in the actual flight of fixed-wing UAVs, including actuator multiplicative and additive faults, thrust saturation cutoff, parameter uncertainties, and external environmental disturbances, are uniformly modeled as cluster uncertainties d. i In order to accurately estimate the uncertainty within a specified time and feed the estimation result back to the control law for fault tolerance compensation, this invention designs a type of time-determined uncertainty observer for each UAV.

[0123] First, define the error:

[0124] Let the speed of the i-th drone be V. i The uncertainty is estimated as Define the estimation error of velocity and uncertainty as:

[0125]

[0126] The goal of the observer is to observe within a finite, specified time t. e Intrinsic drive

[0127] Define the time scaling function again:

[0128] To achieve global convergence within a finite time, a specified time scaling function ξ is constructed.di (t):

[0129]

[0130] Among them, t s T is the time when the observer is enabled. c,i =t e -t s The convergence time is h, where h > 0 is the scaling parameter. It is a positive definite constant. The introduction of this scaling function causes the gain term to vary with time and tend to infinity within a finite time interval, thus ensuring that the error system converges within a specified time.

[0131] S202. Observer update and stability analysis: Design an update law, and ensure that the system achieves accurate estimation within a specified time by combining traditional proportional feedback and dynamic gain terms. The stability of the observer is proved by using the Lyapunov method, ensuring that the error can converge to zero quickly.

[0132] Based on the scaling function of S201 above, the update law of the observer is designed as follows:

[0133]

[0134] Where, α 1i ,α 2i >0,β 1i ,β 2i >4 is the gain for the observer design.

[0135] The structure consists of two parts:

[0136] Part One It uses a traditional proportional feedback term to ensure asymptotic stability;

[0137] Part Two This is a dynamic gain term that varies over time, used to enhance the convergence rate within a preset time period.

[0138] To prove the stability of the above observer, a Lyapunov function is constructed:

[0139]

[0140] Taking its derivative, we get:

[0141]

[0142] Right now:

[0143]

[0144] According to the presupposed time stability theory, when α 2i >0,β2i When the value is greater than 4, there exists a finite time t. e , making Therefore, the designed observer can accurately estimate the uncertainty within a specified time.

[0145] Overall, by using this time-determined uncertainty observer, this invention can obtain an accurate estimate of cluster uncertainty within a finite time, enabling compensation for actuator failures, thrust saturation, and external disturbances, thereby significantly enhancing the controller's fault tolerance and robustness. Furthermore, the proposed time scaling function and dynamic gain design avoid the drawback of traditional observers where convergence rate is affected by initial conditions, allowing the system to quickly enter a stable state during actual flight missions.

[0146] S3. Hierarchical preset performance error transformation: To ensure the predictability of error trajectory performance, preset performance control constraints are applied to both position error and position-velocity composite error, specifically including:

[0147] S301. Set a position error layer, apply a contraction-type performance boundary to the position error, and control the trajectory of the position error through a monotonically decreasing performance function to ensure that the error is always within a preset range:

[0148] Let the desired position of the i-th UAV be... The actual position is p i (t), then its position error is defined as:

[0149]

[0150] To ensure that the position error trajectory always remains within the contraction performance boundary, the following constraints are introduced:

[0151]

[0152] in:

[0153] This is the upper bound of the initial error;

[0154] For a performance function that is monotonically decreasing, the following form is typically chosen:

[0155]

[0156] in, As the initial performance boundary, The boundary value is the steady-state performance boundary, and λ > 0 is the shrinkage rate parameter.

[0157] Further define the error transformation variable as:

[0158]

[0159] At this point, it is only necessary to ensure If it is bounded, then it can be deduced. It always remains within the preset boundaries, thus achieving full-process performance constraints on position error. The position tracking error transformation process is as follows: Figure 4 As shown.

[0160] S302. Velocity Composite Error Layer: Based on position error constraints, velocity error is introduced, and the formation motion coordination is further enhanced through position-velocity composite error. Through hierarchical error transformation, position error and composite error are controlled separately, ensuring that the error remains within the specified performance boundaries throughout the entire flight process. Specifically:

[0161] To further improve the coordination of formation movements, this invention introduces a position-velocity composite error constraint on top of the position error layer.

[0162] Define the speed error as:

[0163]

[0164] in, For the desired speed, V i h i This represents the actual velocity vector of the UAV.

[0165] Further construct the composite error:

[0166] e μi =k p e pi +k v e vi

[0167] Where, k p >0,k v >0 is a design parameter used to balance the contribution of position and velocity errors to the composite error.

[0168] The following performance constraints are imposed on the v-th component:

[0169]

[0170] in, Similar to the location layer, it takes the form of a monotonically decreasing function.

[0171] Further define the composite error transformation variable as:

[0172]

[0173] Similarly, it is only necessary to ensure The boundedness of the position-velocity composite error can be derived from this. The entire process meets the specified performance constraints.

[0174] To ensure that the tracking error of UAV formations not only converges during dynamic flight missions but also meets preset performance constraints throughout the entire process, this invention proposes a hierarchical performance error transformation method. This method introduces specified performance control constraints at two levels: the position error layer and the velocity composite error layer. Through a nonlinear error transformation function, it achieves boundedness and controllable convergence of error variables, thereby ensuring the stability and predictability of the formation's geometry and motion performance.

[0175] By defining the performance error transformation in a hierarchical manner, this invention extends the constraints from single position error to joint position-velocity constraints, ensuring that the UAV not only gradually converges in position error during flight missions, but also enhances velocity synchronization, thus avoiding formation deformation and dispersion caused by velocity differences in traditional methods.

[0176] The advantages of this method lie in its full-process constraint, meaning that the error trajectory remains within the contraction boundary throughout the entire flight process, rather than just meeting the requirements at the convergence moment. It also features a hierarchical coupling design, achieving synchronous control of position and velocity errors through a two-layer error transformation mechanism, thereby significantly improving the overall stability of the formation. Furthermore, this method has enhanced robustness, ensuring the predictability and controllability of the error trajectory even in the presence of actuator failures and external disturbances, thanks to the constraint function.

[0177] S4. Design of fault-tolerant control law based on orientation constraints: The virtual control quantity is converted into thrust and angular velocity commands that satisfy nonholonomic constraints, and the actual execution input is obtained by combining the actuator model.

[0178] After completing system modeling (S1), uncertainty observer design (S2), and hierarchical error transformation (S3), this embodiment further proposes a fault-tolerant control law based on azimuth constraints to map the desired virtual control quantity into actual thrust and angular velocity inputs that satisfy nonholonomic constraints. The core task is to transform the "virtual control command" into thrust and angular velocity executable by the UAV and incorporate a fault-tolerant compensation mechanism. The essence of this control law lies in "localization" and "projection filtering"—global formation stability can be achieved by relying only on neighbor azimuth information.

[0179] This control law fully utilizes neighboring location information, combined with observer output and error transformation results, to ensure the stability and fault tolerance of UAV formations when performing complex maneuvers. Specifically, it includes:

[0180] S401. Application of local orientation information and projection matrix: Each UAV relies only on local orientation information to construct a local projection matrix, ensuring that control signals are projected only in the effective direction and filtering out invalid control inputs;

[0181] First, construct the orientation constraint and local projection matrix:

[0182] Consider a leader-first follower topology where each follower can obtain the relative azimuth information of its neighboring drones. Define the desired azimuth vector:

[0183]

[0184] Where, N i Let represent the set of neighbors of the i-th drone.

[0185] To ensure that the control signal is projected into the azimuth constraint subspace, a local projection matrix is ​​constructed:

[0186]

[0187] in, This represents the actual relative azimuth. This matrix can filter out components in the azimuth direction, ensuring that the error signal is projected into the vertical subspace, effectively avoiding redundant control inputs in unobservable directions.

[0188] S402. Composite Error and Control Law: Design a composite error, combining position and velocity errors to further construct the control law. Specifically, firstly, a composite error is constructed to simultaneously reflect the deviations in position and velocity. The composite error simultaneously characterizes the degree of deviation of the system state in both position and velocity, achieving the effect of dynamically equipping the second-order error with the first-order error, thus providing a unified error signal basis for subsequent control law design.

[0189] Based on the aforementioned composite error, a control law is constructed, which consists of an error feedback term, a feedforward compensation term, and an uncertainty observer compensation term. The error feedback term originates from the composite error and is used to generate feedback control quantities based on the current state deviation, stabilizing the system and gradually eliminating tracking errors. The feedforward compensation term provides advance control based on known prior information about the reference trajectory, offsetting the main control quantities required for tracking under the nominal model, thereby reducing steady-state error and tracking delay. The uncertainty observer compensation term originates from real-time estimation of the unknown parts of the system. It estimates model uncertainties by introducing disturbances or uncertainty observers and generates corresponding compensation signals that are superimposed on the control input to offset the influence of uncertainties on the system, improving control robustness and accuracy. These three terms each perform their respective functions and cooperate with each other: the error feedback term ensures closed-loop stability and basic tracking performance; the feedforward term injects the required control in advance to improve response quality; and the uncertainty compensation term enhances the system's disturbance rejection capability. Together, they form a complete control law.

[0190] Based on the combined error and the transformation error, this invention designs a control law for longitudinal acceleration and angular velocity, the general structure of which is as follows:

[0191] u i =u fb,i +u ff,i +u ob,i

[0192] Among them, u fb,i For the error feedback term, u ff,i For the feedforward compensation term, u ob,i This is the compensation term for the uncertain observer.

[0193] The error feedback term is directly generated from the composite error and is used to ensure the stability of the closed-loop system.

[0194]

[0195] Where K μ This is the positive definite feedback gain matrix. This term originates from Lyapunov stability analysis, ensuring error convergence through negative feedback of the composite error.

[0196] The feedforward compensation term is derived from the dynamic prior information of the desired trajectory:

[0197]

[0198] in Indicates the desired longitudinal acceleration. This represents the desired angular velocity. This term offsets the effect of changes in the reference trajectory on the system, ensuring that the actual state can follow the desired maneuver in advance and avoiding steady-state deviations.

[0199] Considering actuator failure, parameter perturbation, and external disturbances, an uncertainty observer is constructed to estimate the unknowns and obtain the compensation signal:

[0200]

[0201] in For the observer's uncertainty about the cluster d i Real-time estimation. This function aims to offset interference from uncertainties and improve the system's fault tolerance and robustness.

[0202] The longitudinal acceleration and angular velocity control laws ensure the stability of the UAV formation and effectively compensate for uncertainties through error feedback and feedforward compensation. The control laws rely on local neighbor azimuth information, avoiding dependence on the global reference system and reducing computational complexity. Specifically:

[0203] Based on the hierarchical error transformation, this invention further utilizes a local projection matrix to construct a distributed composite tracking error:

[0204]

[0205] Where: k p ,k v >0 represents the design gain;

[0206] γ i This is a normalization factor used to balance the impact of differences in the number of neighbors on the control effect.

[0207] This composite error term includes both the projected differences in position and velocity, and can reflect the dynamic inconsistencies of the formation in the local neighborhood.

[0208] Combined with the output of the uncertain observer and error transformation variables The control law is constructed as follows:

[0209] Longitudinal acceleration input:

[0210]

[0211] Angular velocity input:

[0212]

[0213] Where: k 1i ,k 2i >0 represents the error feedback gain;

[0214] J pi J μi These are the Jacobian matrices for the error transformation, used to realize the transformation from the ε-space to the original error space;

[0215] The derivative of the desired velocity represents the feedforward compensation term.

[0216] The control law consists of two parts:

[0217] Feedback and compensation items Achieve convergence under error constraints;

[0218] Feedforward and Observer Compensation Terms Used to offset the effects of changes in the desired trajectory and uncertainties.

[0219] Since the actual execution input is τ i =sat(τ i,F ), where τ i,F =ρ i τ i,c +φ i Furthermore, by combining uncertainty observer compensation, the errors introduced by multiplicative fault factors, additive bias, and saturation truncation can be significantly reduced, ensuring that the system has fault tolerance.

[0220] Furthermore, the localization design of the control law means that each UAV only needs to rely on local relative information, without the need for global position coordinates or velocity derivatives, thereby reducing sensor dependence and computational complexity and improving the feasibility of the method in actual formation flight.

[0221] The technical effects of this fault-tolerant control law design are mainly reflected in enhanced robustness. Through observer compensation and error projection, the formation can remain stable even under external disturbances and fault conditions. It has fault tolerance guarantees, and the convergence of the system can still be maintained even in the presence of actuator failures and saturation constraints. At the same time, it has low information dependence, and only local azimuth information is needed to complete the control law design, thereby eliminating the dependence on the global reference frame. Furthermore, the performance is guaranteed throughout the entire process. Combined with error transformation design, it can ensure that the formation meets performance constraints throughout the entire flight mission, not just in the steady state phase.

[0222] S5. Stability Analysis and Parameter Condition Proof: After designing the fault-tolerant control law based on orientation constraints, this embodiment further constructs a Lyapunov function to rigorously prove the closed-loop stability of the system and provides the conditions for the value of the control gain to ensure that the formation system can maintain error convergence and fault-tolerant characteristics even in the presence of external disturbances and actuator failures. Specifically, this includes:

[0223] S501. Closed-Loop Error System Modeling and Stability Proof: Based on the aforementioned design, a closed-loop error system is established. The dynamic behavior of position error, velocity error, and combined error is analyzed. By constructing a Lyapunov function, the stability of the system is rigorously proven, ensuring that the error remains within the preset performance boundaries throughout the entire flight process and eventually converges to zero. Specifically:

[0224] First, we model the error system:

[0225] The closed-loop error system, from S1 to S4, mainly includes the following components:

[0226] Position error:

[0227] Speed ​​error:

[0228] Composite error: e μi =k p e pi +k v e vi

[0229] Local projection composite error:

[0230] Combining the control law, the actual closed-loop error dynamics can be obtained as follows:

[0231]

[0232] Among them, the uncertainty d i The S2 observer has obtained an accurate estimate within the specified time.

[0233] To analyze stability, the Lyapunov function is defined as follows:

[0234]

[0235] This function comprehensively considers the changes in velocity error, position error, and composite error, and can fully reflect the degree of deviation of the system.

[0236] Differentiating with respect to V, we have:

[0237]

[0238] Combining the control law expression with error dynamics, we can obtain:

[0239]

[0240] Further utilize the neighbor projection matrix P ij Properties, namely We can obtain:

[0241]

[0242] in:

[0243] The Lyapunov function V measures the degree of deviation of the entire formation system. Its value comprehensively reflects the magnitude of the changes in velocity error, position error, and composite error. As the control law is executed, V decreases monotonically, ensuring the asymptotic stability of the system.

[0244] e vi This represents the speed error of the i-th UAV, i.e., the difference between the actual speed and the expected speed. This term reflects the deviation of the system at the speed level.

[0245] This represents the equivalent error after the position error has been transformed by a preset performance function. Through nonlinear transformation, the original position error is constrained within preset boundaries, ensuring that the entire error convergence process meets performance requirements.

[0246] This represents the equivalent error after transformation of the composite error. The composite error is a weighted combination of position and velocity errors. This transformation is used to ensure that the composite error meets preset performance constraints and avoids overshoot or exceeding limits during convergence.

[0247] P ijIt is an orthogonal projection matrix constructed using neighboring orientation information. Its function is to filter out unobservable radial components and retain only the lateral error components that are related to formation geometry.

[0248] Let represent the minimum eigenvalue of the projection matrix corresponding to the neighbor set of the i-th UAV, ensuring the observability and boundedness of the error under the neighbor constraint.

[0249] This represents the upper bound of the acceleration of the desired trajectory;

[0250] ξ d =sup t ||d i || represents the upper bound of an uncertain quantity.

[0251] S502. Derivation of Gain Condition and Convergence Verification: The sufficient condition for the control gain is derived, proving that the system can still converge stably under actuator failure, thrust saturation, and external disturbances. Through stability analysis and fault tolerance verification, the system's excellent fault tolerance and robustness are ensured. Specifically:

[0252] If the control gain satisfies the following condition:

[0253]

[0254] Then the above disturbance terms can be completely canceled out, and thus we have:

[0255]

[0256] According to Lyapunov's stability theorem:

[0257] 1. All error variables are uniformly bounded;

[0258] 2. Further combining the error transformation relationship, we can obtain the position error e pi and speed error e vi The original error, which has not yet undergone nonlinear error transformation, remains within the preset performance boundary throughout the entire process;

[0259] 3. The system is asymptotically stable under the specified performance constraints.

[0260] Through the stability analysis described above, this invention demonstrates that the designed fault-tolerant control law can not only effectively resist the influence of actuator failures and external disturbances, but also ensure that the formation error converges to zero throughout the entire flight mission, strictly adhering to the preset performance boundary conditions. Compared with traditional methods that only guarantee asymptotic convergence, this invention has significant advantages in terms of performance guarantee, as the error is controlled throughout the entire process rather than only meeting performance requirements at the convergence endpoint; it also possesses fault tolerance, effectively offsetting actuator failures, thrust saturation, and external disturbances through observer compensation and gain constraints; in terms of robustness, the system can still maintain stable convergence even with parameter perturbations and environmental uncertainties; furthermore, this invention has distributed implementation characteristics, relying only on local neighbor azimuth information to guarantee the stability of the global formation, thereby significantly improving the feasibility of engineering applications.

[0261] The following simulation examples demonstrate the performance-tolerant formation control method for distributed fixed-wing UAVs based on azimuth information described in this invention.

[0262] This embodiment uses a simulation platform for numerical simulation, with an integration step size of 0.01s and a total simulation time of 60s. The table below shows the dynamic parameters, control gain, and observer parameters of the UAV.

[0263] Parameter categories Parameter symbol numerical values illustrate drone quality <![CDATA[m i ]]> 1.5kg All drones are the same Thrust Limit <![CDATA[τ max ]]> 12N Actuator saturation limit Thrust Lower Limit <![CDATA[τ min ]]> -2N Actuator saturation lower limit initial velocity <![CDATA[V i (0)]]> 8m / s Initial speed of horizontal flight Control gain <![CDATA[k p ,k v ]]> 1.5,2.0 Position and velocity weighting coefficients Error boundary <![CDATA[M pi ,M μi ]]> 5.0,3.0 Initial constraint boundary Shrinkage parameters λ 0.5 Performance boundary contraction rate Observer Gain <![CDATA[α 1i ,a 2i ]]> 2.0,2.5 Guarantee convergence within the specified time Scaling function parameters h 0.8 Specified time scaling function parameters

[0264] In one specific embodiment, the formation consists of eight fixed-wing UAVs, including one leader, one first follower, and six ordinary followers. Their initial positions are set as follows:

[0265] Leader: p1(0) = [0,0,0] T ;

[0266] First follower: p2(0) = [10,0,0] T ;

[0267] Other followers: p3(0) = [0, 10, 0] T p4(0) = [10, 10, 0] T p5(0) = [0, 0, 10] T p6(0) = [0, 10, 10] T p7(0) = [10, 10, 10] T p8(0) = [10, 0, 10] T .

[0268] All UAVs have an initial velocity of 8 m / s and an initial heading vector of [1, 0, 0]. TThe desired formation task is to translate along the X-axis, while gradually shrinking the formation size to half of its original size throughout the mission.

[0269] To verify the fault tolerance and robustness of the method of the present invention, the following faults and disturbances were set:

[0270] At t=15s, the third UAV experienced a multiplicative actuator failure, and the thrust efficiency decreased to ρ3=0.7;

[0271] At t=20s, the fourth UAV experienced an additive actuator failure, resulting in a constant thrust bias φ4=2.0N;

[0272] All UAV thrust is limited to the saturation constraint range [-2, 12] N;

[0273] The external disturbance is set to a bounded sinusoidal signal:

[0274] d i (t)=0.5sin(0.2t), i=1,...,6.

[0275] To highlight the advantages of the method of this invention, three comparative methods were designed: Method 1 is the method of this invention, which uses a time uncertainty observer, hierarchical error transformation, and a fault-tolerant control law based on orientation constraints; Method 2 is traditional consistency control, which does not introduce error constraints and fault-tolerant mechanisms, and only relies on local relative position feedback; Method 3 is a robust control method, which suppresses disturbances through fixed gain, but does not have a specified performance guarantee.

[0276] Figure 5 The trajectory evolution of the eight UAVs in this embodiment is shown. Under the method of this invention, the formation converges quickly and remains stable with a smooth trajectory; the traditional method shows a significant deviation after a fault occurs; the robust method can maintain stability but has a large error. Figure 6 The paper demonstrates the change in position tracking error of a typical follower in three dimensions over time. The method of this invention can guarantee that the error always stays within the specified performance boundaries and converges to zero, while traditional and robust methods both exhibit error out-of-bounds errors. Figure 7 The combined tracking error indicates that the speed is synchronized and constrained, and both traditional and robust methods exhibit speed oscillations. Figure 8 The control input τ in this invention i The curve represents the constraint boundary of the control input, where the dashed line indicates the constraint boundary. It can be seen that the control input is always limited within a preset range, and reaches the boundary value during the formation stage due to a large tracking error. Figure 9 It is the lumped uncertainty estimation error d i The curves show the estimation performance of the observer for uncertainties within a specified time. The method of this invention can accurately converge to the true value within a finite time, significantly improving the fault tolerance and compensation capability.

[0277] Simulation results fully verify the superior performance of the proposed method: even under various uncertainties such as actuator multiplicative faults, additive faults, thrust saturation, and external disturbances, the proposed method ensures that position and velocity errors are always constrained by the specified performance boundaries throughout the mission and eventually converge to zero, demonstrating the advantage of controllable performance throughout the entire process. By introducing a specified time uncertainty observer, the effects of faults and disturbances are effectively compensated, achieving a balance between fault tolerance and robustness. Simultaneously, the control law relies only on local neighbor azimuth information, reducing dependence on the global reference frame and high-precision sensors, thus improving the method's practicality and feasibility. Furthermore, the method exhibits good scalability and can be directly extended to larger-scale UAV swarms or other types of multi-agent systems while maintaining stability and formation performance, demonstrating broad engineering application prospects.

[0278] This invention proposes a pre-defined performance formation control method based on azimuth information for fixed-wing unmanned aerial vehicles (UAVs) facing actuator failures and system uncertainties. The UAV dynamics model is reconstructed as a nonholonomic second-order model, explicitly considering actuator efficiency losses, bias faults, input saturation, and environmental disturbances. Under the leader-first follower communication structure assumption, this invention establishes a hierarchical fault-tolerant control framework, including a time-bound observer for uncertainty estimation, a hierarchical error transformation to guarantee performance constraints, and a distributed azimuth-based controller for maneuvering formation tracking. Through rigorous stability analysis and comparative simulation verification, the method of this invention outperforms existing methods in both convergence performance and fault tolerance. Its key advantages are: employing a more realistic UAV dynamics model, ensuring pre-defined performance tracking throughout the process through hierarchical error transformation, and effectively integrating pre-defined performance control into the azimuth-based control framework, achieving stable and efficient formation control even with information loss due to projection calculations.

[0279] The foregoing has provided a detailed description of a distributed fixed-wing UAV pre-set performance fault-tolerant formation control method based on azimuth information provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention, and the descriptions of the embodiments above are only for the purpose of helping to understand the core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A distributed fixed-wing unmanned aerial vehicle (UAV) pre-set performance fault-tolerant formation control method, characterized in that: Includes the following steps: S1. System Modeling and Problem Description: Establish a nonholonomic dynamic model of the fixed-wing UAV, considering uncertainties such as actuator failure, thrust saturation, and environmental disturbances, and unify them into cluster uncertainties; S2. Uncertainty Observer Design: For the introduced cluster uncertainty, a time-defined observer is established, and a time scaling function is used to establish the update law for uncertainty estimation to ensure accurate estimation within the specified time. S3. Hierarchical error transformation: Set the position error and set the position-velocity composite error on this basis. Apply control constraints to the position error and the position-velocity composite error. Control the position error trajectory through a monotonically decreasing performance function to ensure that the error is always within the preset performance boundary. S4. Design of fault-tolerant control law based on orientation constraint: By using a local projection matrix, the error signal is projected into the orientation constraint subspace, thereby filtering out invalid control inputs. The control law is constructed in combination with the error, and the stability of the UAV formation is ensured through error feedback and feedforward compensation.

2. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: Also includes: S5. Stability Analysis and Parameter Condition Proof: The closed-loop stability of the system is proved using the Lyapunov method, and the conditions for the value of the control gain are given to ensure that the formation can still operate stably under external disturbances and actuator failures.

3. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: The nonholonomic dynamics model of the fixed-wing UAV includes the kinematic and dynamic equations of each UAV, describing the UAV's flight path angle, heading angle, mass, drag, environmental disturbances, and control inputs.

4. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: The cluster uncertainty sets characterize the combined effects of actuator multiplicative faults, additive biases, thrust saturation errors, parameter perturbations, and external disturbances.

5. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: S1 specifically includes: S101. Establish a dynamic model of a fixed-wing UAV. Each UAV is modeled as a rigid body, and its motion is described by both the inertial coordinate system and the body coordinate system. The kinematic and dynamic equations of the i-th UAV are as follows: Where, p i =[x i ,y i ,z i ] T V represents the position vector. i Indicates airspeed, γ i and ψ i These are the flight path angle and heading angle, respectively. For actual quality, D i As resistance, d ei For environmental interference, U i L i and μ i These are the control inputs for thrust, lift, and roll angle, respectively. S102. Set the formation topology as the first follower, and the other followers establish perception with local neighbors through orientation information, and construct an orthogonal projection matrix using local orientation information.

6. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 5, characterized in that: In the topology, the first follower only perceives the leader, while other followers perceive the neighbors with smaller numbers, ensuring that the orientation constraints are not collinear and that all orientation constraints conform to the center-to-scale geometric description.

7. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: The specified time observer ensures that the system error converges within a finite time by introducing a time scaling function and a dynamic gain term.

8. The distributed fixed-wing UAV preset performance fault-tolerant formation control method according to claim 1, characterized in that: The control law combines error feedback and feedforward compensation to ensure the stability and fault tolerance of the formation when performing complex maneuvers. It does not rely on the global coordinate system, but only on the local neighbor azimuth information.

9. A distributed fixed-wing UAV pre-set performance fault-tolerant formation control method according to claim 1, characterized in that: The update law for the time-defined observer is: Where, α 1i ,α 2i >0,β 1i ,β 2i >4 is the gain for the observer design.