Cluster fault-tolerant consistency control method for heterogeneous unmanned system

Abstract

The invention discloses a cluster fault-tolerant consistency control method for a heterogeneous unmanned system, and relates to the technical field of unmanned system cross-domain cooperative control. The method comprises the following steps: firstly, establishing a collaborative motion control model of a heterogeneous unmanned system consisting of a group of unmanned ships and unmanned aerial vehicles; then, based on a newly developed error conversion mechanism capable of avoiding singularity, a distributed reference state estimator capable of meeting preset performance requirements is designed to obtain an output trajectory of a cluster leader. Secondly, a neural network approximator is used for processing actuator faults of the system, and a new weight updating law is designed for the neural network approximator through the immersion and invariant self-adaption theory; and finally, providing a cluster fault-tolerant consistency control strategy with preset time and predefined precision in combination with dynamic surface control and a newly developed error conversion mechanism to realize cluster consistency of the heterogeneous unmanned system under the actuator fault. The method has good practical significance and application prospects in the field of heterogeneous unmanned system cross-domain cooperation and fault-tolerant control.

Description

Technical Field

[0001] The present invention relates to the technical field of cross-domain collaborative control of unmanned systems, and in particular to a cluster fault-tolerant consistency control method for heterogeneous unmanned systems. Background Art

[0002] Heterogeneous unmanned systems have received extensive attention and research from industry and academia due to their efficient execution capabilities and huge application potential. Compared with homogeneous unmanned systems, heterogeneous unmanned systems are composed of multiple homogeneous or heterogeneous unmanned equipment with different parameters and performance. Through cross-domain perception, cluster collaboration and autonomous decision-making, the capabilities of all aspects of the system can be maximized, thus providing strong technical support for smart people's livelihood.

[0003] However, achieving cluster consistency control of heterogeneous unmanned systems is currently a challenging problem, which requires that the unmanned equipment in the system evolve into several different sub-clusters, and the state of each sub-cluster is consistent. On the other hand, with the increase in the scale and number of components of heterogeneous unmanned system clusters, fault tolerance is crucial to maintaining overall stability and reliable coordinated motion when a sub-member of the system fails. Therefore, there is an urgent need to develop and research efficient heterogeneous unmanned system fault-tolerant control technology. In addition, considering that the cluster consistency of heterogeneous unmanned systems has pre-defined transient and steady-state behaviors is also a very meaningful study at present, which reflects the inevitable requirement for the development of system control to a higher level. Therefore, in order to enhance the safety and reliability of the collaborative operation of heterogeneous unmanned system clusters and optimize the control effect and performance of existing technologies and solutions, this application studies a cluster fault-tolerant consistency control method for heterogeneous unmanned systems. Summary of the invention

[0004] In order to solve the above technical problems and realize cluster consistency control of heterogeneous unmanned systems with actuator failures, the purpose of the present invention is to design a cluster fault-tolerant consistency control method for heterogeneous unmanned systems to ensure that in the event of a failure, the unmanned equipment in the heterogeneous unmanned system can also simultaneously track the trajectories of different types of cluster leaders, and the formation error can converge to within a preset limit within a specified time.

[0005] Technical Solution

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] First, a unified collaborative motion control model for heterogeneous unmanned systems under actuator failure is constructed. Secondly, an error conversion mechanism that can avoid singularities is designed to achieve performance-constrained control of the system. Then, on this basis, a distributed reference state estimator that can meet the preset performance requirements is developed to estimate the output information of the cluster leader. Finally, under the framework of dynamic surface control, combining the above error conversion mechanism with the neural network adaptive approximation theory, a cluster fault-tolerant consistency control strategy with preset time and predefined accuracy is designed to achieve cross-domain collaboration and fault tolerance of heterogeneous unmanned systems under actuator failure.

[0008] The control method comprises the following steps:

[0009] Step 1: For a heterogeneous unmanned system consisting of a group of N1 unmanned boats and N2 unmanned aerial vehicles, a unified collaborative motion control model of the heterogeneous unmanned system with actuator failure is established. The specific process is as follows.

[0010] The first step is to establish the dynamic model of the i-th unmanned boat, as shown below:

[0011]

[0012] in,

[0013]

[0014] is the Jacobi transformation matrix, η i =[x i ,y i ,ψ i ] T is the position and heading of the i-th unmanned boat in the inertial coordinate system, v i =[u i ,v i ,r i ] T is the sway, drift and pitch speed of the i-th unmanned boat in the attached coordinate system, and are η i and v i The first derivative of i is the control input, M i , C i (v i ) and D i (v i ) are the inertia matrix, Coriolis centripetal force matrix and damping matrix respectively. Define x i1 =η i 、x i2 =v i and u i =τi , formula (1) can be rewritten as:

[0015]

[0016] Among them, f i1 =J(ψ i )v i -v i and is a nonlinear term, g i1 =I3 and is the gain term, and I3 represents the 3×3 unit matrix.

[0017] The second step is to establish the dynamic model of the i-th fixed-wing UAV as follows:

[0018]

[0019] Among them, x i ,y i 、z i is the position of the i-th fixed-wing UAV in the inertial coordinate system, and x i ,y i and z i The derivative of V i is the speed of the i-th fixed-wing UAV, V i The derivative of and are the heading angle and track angle of the i-th fixed-wing UAV, respectively. and They are respectively i and γ i The derivative of m i is the mass of the i-th fixed-wing UAV; g is the acceleration due to gravity; μ i , α i and β i are the inclination angle, attack angle and sideslip angle of the i-th fixed-wing UAV respectively; T i , D i , Y i and L i are respectively the thrust, drag, lateral force and lift exerted on the i-th fixed-wing UAV, and the specific expressions are as follows:

[0020]

[0021] Among them, C iL , C iD and C iY They are lift coefficient, drag coefficient and force coefficient respectively, and the specific forms are as follows:

[0022]

[0023] Among them, T imax Indicates the maximum thrust of the engine; For thrust throttle settings; represents dynamic pressure, ρ0 represents air density; s i is the wing area; C iL0 , C iLα , C iD0 , C iDα , C iDα2 , C iY0 , C iYβ is the aerodynamic coefficient. Define x i1 =[x i ,y i ,z i ] T , x i2 =[V i ,γ i ,γ i ] T ,u i =[δ Ti ,α i sinμ i ,α i cosμ i ] T , formula (3) can be rewritten as:

[0024]

[0025] Among them, g i1 =I3 and is the gain term, f i1 =[f i11 ,f i12 ,f i13 ] T and f i2 =[f i21 ,f i22 ,f i23 ] T is a nonlinear term, which is specifically expressed as follows:

[0026] f i11 =V i cosγ i cos i -V i

[0027] f i12 =V i cosγ i sinx i -xi

[0028] f i13 =V i sinγ i -γ i

[0029] f i21 =-D i / m i -gsinγ i

[0030]

[0031] In the third step, consider the following actuator failure model:

[0032] u i =ρ i u i0 +b i (5)

[0033] Among them, ρ i =diag{ρ i1 ,ρ i2 ,ρ i3} is the actuator efficiency loss factor, which is a diagonal matrix satisfying ρ i1 ,ρ i2 ,ρ i3 ∈(0,1]; b i is a bounded deviation fault vector; u i0 It is the command control input signal.

[0034] In the fourth step, substituting equation (5) into equation (2) and equation (4), we can obtain the unified cooperative motion control model of heterogeneous unmanned systems under actuator failure:

[0035]

[0036] Among them, h i =g i2 [(ρ i -I3)u i0 +b i ] is the lumped fault item.

[0037] Step 2: To achieve the performance constraints of the system, the following error conversion mechanism is constructed to avoid singularity:

[0038]

[0039] Where, π(t) = 1-β(t) 2 z(t) 2, z(t) represents the error variable of the system, β(t) is a time-varying scalar function with the following properties:

[0040] 1) β(t) is at least C on the interval [t0,∞ n+1 Class function, n is the system order;

[0041] 2) β(t) is a continuous non-decreasing function, with β(t0) = 0 and where t p <∞ is a preset time constant;

[0042] 3)

[0043] 4) For t>t0+t p ,have and Among them, 0<ζ<∞ and 0<t p <∞ are the predefined accuracy and preset time respectively. Further, taking the derivative of s(t) yields:

[0044]

[0045] in, In addition, in order to facilitate the design of the subsequent distributed reference state estimator, the following transformations are performed:

[0046]

[0047] in, And from the properties of β(t), we know is bounded. For the convenience of description below, (t) is omitted.

[0048] Step 3: In order to estimate the output information of the cluster leader, based on the error conversion mechanism proposed in step 2, a distributed reference state estimator that can meet the preset performance requirements is developed as follows:

[0049]

[0050] Among them, there are:

[0051]

[0052] in, and ι2 are control parameters, and Representation Matrix The maximum eigenvalue of Representation Matrix The minimum eigenvalue of μ. i01 =diag{μ i01q}, q=1,2,3 is a 3×3 diagonal matrix, satisfying satisfy is the output of the above distributed reference state estimator. express y i The estimate, y i is a constant and satisfies and y i Respectively represent y i Using the error conversion mechanism mentioned above, we can get s i0 The dispersion expression of is:

[0053]

[0054] in, β0 is a time-varying scalar function defined by the user, z i0 =[z i01 ,z i02 ,z i03 ] T is the cluster consistency tracking error, defined as follows:

[0055]

[0056] in, is the output of the leader of the cluster where the i-th system member is located. b i =1 means that the i-th system member is a fixed node and can receive the output information of the cluster leader. The selection method of fixed nodes in the system is as follows: According to the multi-task clustering requirements, the directed graph The topological nodes in are classified into disjoint task sets For i≤N=N1+N2, if d i = 0, then b i =1; if d i ≠0 and there is Then b i =1; if d i ≠0 and there is Then b i =1.

[0057] Step 4: After obtaining the output information of the leader using the distributed reference state estimator designed in step 3, under the framework of dynamic surface control, based on the error conversion mechanism and neural network adaptive approximation theory proposed in step 2, a cluster fault-tolerant consistency control strategy with preset time and predefined accuracy is designed to achieve cross-domain collaboration and fault tolerance of heterogeneous unmanned systems under actuator failure. The specific process is as follows:

[0058] Step 1: Define the system's position tracking error zi1 , and design virtual control law

[0059] definition is the position tracking error of the i-th system member, where x id represents the desired cluster formation configuration. Using the error conversion mechanism mentioned above, we can get:

[0060]

[0061] in, β1 is a time-varying scalar function defined by the user, s i1 =[s i11 ,s i12 ,s i13 ] T . Further, by taking the derivative of the above formula (14), we can get:

[0062]

[0063] in, Indicates z i1q The derivative of . Combining equations (6) and (10), the compact expression of equation (15) is:

[0064]

[0065] Among them, μ i11 =diag{μi i111 ,μ i112 ,μ i113},μ i02 =[μ i121 ,μi i122 ,μ i123 ] T , is x id The derivative of .

[0066] Next, design a virtual control law for system (6): for:

[0067]

[0068] Among them, k i1 is a control parameter, is μ i11 The derivative of .

[0069] The command filter is designed as:

[0070]

[0071] Among them, τ i is the filter error, is the filter output, for The derivative of . Then, it is defined as is the filtering error of formula (18), The derivative is:

[0072]

[0073] in,

[0074] Step 2: Define the system's velocity tracking error s i2 , and design the actual control law u i0 .

[0075] definition is the velocity tracking error of the i-th system member, along the system (6) for s i2 The derivative is:

[0076]

[0077] For system (6), the actual control law u i0 Designed for:

[0078]

[0079] Among them, k i2 To control the parameters, is the concentrated fault item h i The estimated value of .

[0080] Step 3: Estimate h based on the approximation ability of the neural network i , and introduces immersion and invariant adaptation theory to design a new weight update law.

[0081] Using neural network to approximate the lumped fault term h in system (6) i , we can get:

[0082]

[0083] Among them, W i * represents the ideal weight vector, is the Gaussian basis function vector, is the approximate error, satisfying ε a is a constant matrix. Since the ideal weight vector W i * is unknown but bounded, so we use immersion and invariant adaptation theory to obtain W i *Based on this, the approximate value of formula (22) can be expressed as:

[0084]

[0085] in, W i * For the convenience of expression, we will use θ i , Φ i and ε i express and

[0086] definition is the estimated error of the ideal weight. Along the system (6), s i3 The derivative of can be written as:

[0087]

[0088] in, Represents θ i right Next, using the immersion and invariant adaptation theory, we design The update law and function θ i , as shown below:

[0089]

[0090] Among them, k i3 is the control parameter. Let θ ipq Represents the matrix θ i The (p,q)th element in , we can get:

[0091]

[0092] Among them, Φ ip Represents Φ i The pth element in .

[0093] In summary, the distributed reference state estimator composed of equations (11)-(12), the heterogeneous unmanned system control law composed of equations (16)-(18), and the immersed and invariant adaptive neural network estimator composed of equations (20)-(22) are combined to obtain a cluster fault-tolerant consistency control scheme with preset time and predefined accuracy for heterogeneous unmanned systems, thereby realizing cross-domain cluster collaborative control and fault tolerance of heterogeneous unmanned systems with preset performance in the event of actuator failure.

[0094] The beneficial effects of the present invention are:

[0095] (1) The present invention discloses a cluster fault-tolerant consistency control method for heterogeneous unmanned systems. The method takes into account the cluster coordination and fault tolerance problems of heterogeneous unmanned systems affected by actuator failures. The designed method can not only ensure the stability of cluster consistency control of heterogeneous unmanned systems under actuator failures, but also enable the unmanned equipment in the system to simultaneously track the trajectories of different types of cluster leaders.

[0096] (2) A new error conversion mechanism that can avoid singularity is designed to achieve performance-constrained control of system errors. Compared with traditional preset performance control, the designed error conversion mechanism relaxes the constraints on the availability of initial error values.

[0097] (3) Utilizing the designed error conversion mechanism and pinning control idea, a distributed reference state estimator is developed that can achieve the user's preset performance requirements to obtain an estimate of the cluster leader's output information.

[0098] (4) A neural network is used to approximate the lumped fault terms of heterogeneous unmanned systems, and the immersion and invariant adaptation theory is introduced to design a new weight update law for the neural network.

[0099] (5) Based on dynamic surface control and the designed error conversion mechanism, a cluster fault-tolerant consistency control strategy with preset time and predefined accuracy is proposed to solve the cluster consistency problem of heterogeneous unmanned systems under actuator failure.

[0100] (6) The present invention simultaneously takes into account the cluster consistency, fault tolerance and performance constraint requirements of heterogeneous unmanned systems, and has great practical significance and application prospects in the field of cross-domain collaborative control of heterogeneous unmanned systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] Figure 1 It is a flow chart of the cluster fault-tolerant consistency control scheme for heterogeneous unmanned systems under actuator failures;

[0102] Figure 2 This is a system block diagram of the cluster fault-tolerant consistency control scheme for heterogeneous unmanned systems under actuator failures;

[0103] Figure 3 The communication topology diagram for heterogeneous unmanned systems;

[0104] Figure 4 is the output of the distributed reference state estimator Graphs;

[0105] Figure 5 is the cluster consistency error z of the heterogeneous unmanned system i0 Time response curve of

[0106] Figure 6is the position tracking error z of the heterogeneous unmanned system i1 Time response curve of

[0107] Figure 7 is the aggregate fault term h of the heterogeneous unmanned system i and its estimated value Graph.

[0108] Figure 8 The control input u of the heterogeneous unmanned system is i0 Graph. DETAILED DESCRIPTION

[0109] The control method of the present invention is further explained in conjunction with the attached diagrams.

[0110] (a) For a heterogeneous unmanned system consisting of a group of N1 unmanned boats and N2 unmanned aerial vehicles, a unified collaborative motion control model for the heterogeneous unmanned system including actuator failures is established. The specific process is as follows.

[0111] First, the dynamic model of the i-th unmanned boat is established as follows:

[0112]

[0113] in,

[0114]

[0115] is the Jacobi transformation matrix, η i =[x i ,y i ,ψ i ] T is the position and heading of the i-th unmanned boat in the inertial coordinate system, v i =[u i ,v i ,r i ] T is the sway, drift and pitch speed of the i-th unmanned boat in the attached coordinate system, and are η i and v i The first derivative of i is the control input, M i , C i (v i ) and D i (v i ) are the inertia matrix, Coriolis centripetal force matrix and damping matrix respectively. i , C i (v i ) and D i (vi ) is expressed as follows:

[0116]

[0117] Define x i1 =η i 、x i2 =v i and u i =τ i , formula (1) can be rewritten as:

[0118]

[0119] Among them, f i1 =J(ψ i )v i -v i and is a nonlinear term, g i1 =I3 and is the gain term, and I3 represents the 3×3 unit matrix.

[0120] Secondly, the dynamic model of the i-th fixed-wing UAV is established as follows:

[0121]

[0122] Among them, x i ,y i 、z i is the position of the i-th fixed-wing UAV in the inertial coordinate system, and x i ,y i and z i The derivative of V i is the speed of the i-th fixed-wing UAV, V i The derivative of and are the heading angle and track angle of the i-th fixed-wing UAV, respectively. and They are respectively i and γ i The derivative of m i is the mass of the i-th fixed-wing UAV; g is the acceleration due to gravity; μ i , α i and β i are the inclination angle, attack angle and sideslip angle of the i-th fixed-wing UAV respectively; T i , D i , Y i and L iare respectively the thrust, drag, lateral force and lift exerted on the i-th fixed-wing UAV, and the specific expressions are as follows:

[0123]

[0124] Among them, C iL , C iD and C iY They are lift coefficient, drag coefficient and force coefficient respectively, and the specific forms are as follows:

[0125]

[0126] Among them, T imax Indicates the maximum thrust of the engine; For thrust throttle settings; represents dynamic pressure, ρ0 represents air density; s i is the wing area; C iL0 , C iLα , C iD0 , C iDα , C iDα2 , C iY0 , C iYβ is the aerodynamic coefficient. Define x i1 =[x i ,y i ,z i ] T , x i2 =[V i ,γ i ,γ i ] T ,u i =[δ Ti ,α i sinμ i ,α i cosμ i ] T , formula (3) can be rewritten as:

[0127]

[0128] Among them, g i1 =I3 and is the gain term, f i1 =[f i11 ,f i12 ,f i13 ] T and f i2 =[f i21 ,f i22 ,f i23 ] T is a nonlinear term, which is specifically expressed as follows:

[0129] f i11 =V i cosγ i cosX i -V i

[0130] f i12 =V i cosγ i sinx i -x i

[0131] f i13 =V i sinγ i -γ i

[0132] f i21 =-D i / m i -gsinγ i

[0133]

[0134] Then, consider the following actuator failure model:

[0135] u i =ρ i u i0 +b i (5)

[0136] Among them, ρ i =diag{ρ i1 ,ρ i2 ,ρ i3} is the actuator efficiency loss factor, which is a diagonal matrix satisfying ρ i1 ,ρ i2 ,ρ i3 ∈(0,1]; b i is a bounded deviation fault vector; u i0 It is the command control input signal.

[0137] Finally, substituting equation (5) into equation (2) and equation (4), we can obtain the unified cooperative motion control model of heterogeneous unmanned systems under actuator failure:

[0138]

[0139] Among them, h i =g i2 [(ρ i -I3)u i0 +b i ] is the lumped fault item.

[0140] (b) In order to achieve the performance constraints of the system, the following error conversion mechanism is constructed to avoid singularity:

[0141]

[0142] Where, π(t) = 1-βt) 2 z(t) 2 , z(t) represents the error variable of the system, β(t) is a time-varying scalar function with the following properties:

[0143] 1) β(t) is at least C on the interval [t0,∞ n+1 Class function, n is the system order;

[0144] 2) β(t) is a continuous non-decreasing function, with β(t0) = 0 and where t p <∞ is a preset time constant;

[0145] 3)

[0146] 4) For t>t0+t p ,have and Among them, 0<ζ<∞ and 0<t p <∞ are the predefined accuracy and preset time respectively. Further, taking the derivative of s(t) yields:

[0147]

[0148] in, In addition, in order to facilitate the design of the subsequent distributed reference state estimator, the following transformations are performed:

[0149]

[0150] in, And from the properties of β(t), we know is bounded. For the convenience of description below, (t) is omitted.

[0151] (c) In order to estimate the output trajectory of the cluster leader, the error conversion mechanism proposed in (b) is adopted to develop a distributed reference state estimator that can meet the user's preset performance requirements, as shown below:

[0152]

[0153] Among them, there are:

[0154]

[0155] Among them, and ι2 are control parameters, and Representation Matrix The maximum eigenvalue of Representation Matrix The minimum eigenvalue of μ. i01 =diag{μi i01q}, q=1,2,3 is a 3×3 diagonal matrix, satisfying satisfy is the output of the above distributed reference state estimator. express y i The estimate, y i is a constant and satisfies and y i Respectively represent y i Using the error conversion mechanism mentioned above, we can get s i0 The dispersion expression of is:

[0156]

[0157] in, β0 is a time-varying scalar function defined by the user, z i0 =[z i01 ,z i02 ,z i03 ] T is the cluster consistency tracking error of the i-th system member, defined as follows:

[0158]

[0159] in, is the output of the leader of the cluster where the i-th system member is located. b i =1 means that the i-th system member is a fixed node and can receive the output information of the cluster leader. The selection method of fixed nodes in the system is as follows: According to the multi-task clustering requirements, the directed graph The topological nodes in are classified into disjoint task sets For i≤N=N1+N2, if Then b i =1; if And there is Then b i =1; if And there is Then b i =1. Define the output error of the estimator as Then the generalized form of formula (13) is:

[0160]

[0161] in, represents the Kroll product, is the Laplace matrix, is the degree matrix, is the adjacency matrix, is an N×N diagonal matrix, N=N1+N2,

[0162] Then, the stability analysis of the proposed distributed reference state estimator (10)-(11) is performed. First, the derivative of equation (12) yields:

[0163]

[0164] By writing formula (15) in a compact form, we can obtain:

[0165]

[0166] Combining formula (14), we have:

[0167]

[0168] in, express Secondly, along equations (10)-(11), calculate the output error of the estimator The derivative of is:

[0169]

[0170] in, y i for y i The lower bound of . Following formula (11), we calculate The derivative of is:

[0171]

[0172] Further, the compact form of equation (18) and equation (19) is:

[0173]

[0174] in,

[0175] Consider the following candidate Lyapunov function:

[0176]

[0177] in, Following equations (17), (20)-(21), taking the derivative of L0, we can obtain:

[0178]

[0179] in, Control parameter k0, The choice of ι2 should satisfy To ensure that the error signal is bounded.

[0180] Therefore, the designed distributed reference state estimator can estimate the output trajectory of the leader with a predefined error tracking accuracy within the time preset by the user.

[0181] (d) After obtaining the output information of the leader using the distributed reference state estimator designed in step 3, under the framework of dynamic surface control, based on the error conversion mechanism and neural network adaptive approximation theory proposed in step 2, a cluster fault-tolerant consistency control strategy with preset time and predefined accuracy is designed to achieve cross-domain collaboration and fault tolerance of heterogeneous unmanned systems under actuator failure. The specific process is as follows:

[0182] First, define the position tracking error z of the system i1 , and design virtual control law

[0183] definition is the position tracking error of the i-th system member, where x id represents the desired cluster formation configuration. Using the error conversion mechanism mentioned above, we can get:

[0184]

[0185] in, β1 is a time-varying scalar function defined by the user, s i1 =s i11 ,s i12 ,s i13 ] T . Further, by taking the derivative of the above formula (14), we can get:

[0186]

[0187] in, Indicates z i1q The derivative of . Combining equations (6) and (10), the compact expression of equation (15) is:

[0188]

[0189] Among them, μ i11 =diag{μi111 ,μ i112 ,μ i113},μ i02 =[μ i121 ,μ i122 ,μ i123 ] T , is x id The derivative of .

[0190] Next, design a virtual control law for system (6): for:

[0191]

[0192] Among them, k i1 is a control parameter, is μ i11 The derivative of .

[0193] The command filter is designed as:

[0194]

[0195] Among them, τ i is the filter error, is the filter output, for The derivative of . Then, it is defined as is the filtering error of formula (18), The derivative is:

[0196]

[0197] in,

[0198] Consider the following candidate Lyapunov function

[0199]

[0200] According to equations (26), (27) and (29), we can obtain L i1 The derivative of , we can get:

[0201]

[0202] Secondly, define the velocity tracking error s of the system i2 , and design the actual control law u i0 .

[0203] definition is the velocity tracking error of the i-th system member, along the system (6) for s i2 The derivative is:

[0204]

[0205] For system (6), the actual control law u i0 Designed for:

[0206]

[0207] Among them, k i2 To control the parameters, is the concentrated fault item h i The estimated value of .

[0208] Consider the following candidate Lyapunov function

[0209]

[0210] According to equations (32) and (33), we can obtain L i2 The derivative of , we can get:

[0211]

[0212] in, represents the estimation error of the lumped fault term.

[0213] Then, h is estimated based on the approximation ability of the neural network. i , and introduces the immersion and invariant adaptation theory to design a new weight update law to obtain

[0214] The lumped fault term h in system (6) i can be approximated by a neural network with arbitrary precision, specifically expressed as:

[0215]

[0216] Among them, W i * represents the ideal weight vector, is the Gaussian basis function vector, is the approximate error, satisfying ε a is a constant matrix. Since the ideal weight vector W i * is unknown but bounded, so we use immersion and invariant adaptation theory to obtain W i * Based on this, the approximate value of formula (36) can be expressed as:

[0217]

[0218] in, W i* For the convenience of expression, we will use θ i , Φ i and Φ i express and

[0219] definition is the estimated error of the ideal weight. Along the system (6), s i3 The derivative of can be written as:

[0220]

[0221] in, Represents θ i right Next, using the immersion and invariant adaptation theory, we design The update law and function θ i , as shown below:

[0222]

[0223] Among them, k i3 is the control parameter. Let θ ipq Represents the matrix θ i The (p,q)th element in , we can get:

[0224]

[0225] Among them, Φ ip Represents Φ i The pth element in .

[0226] Consider the following candidate Lyapunov function

[0227]

[0228] Where tr{·} represents the trace of the matrix. i3 The derivative of , we can get:

[0229]

[0230] Stability analysis: Consider the following candidate Lyapunov function

[0231] L i =L i1 +L i2 +L i3 (43) where tr{·} represents the trace of the matrix. Using equations (31), (35) and (42), we can obtain L i The derivative of , we can get:

[0232]

[0233] in, Control parameters are selected to satisfy

[0234] Through stability analysis, it is proved that the proposed control method can make the tracking error of the fault heterogeneous unmanned system converge to the specified accuracy range within a finite time.

[0235] (e) The above-mentioned distributed reference state estimator, neural network and weight adaptive law, cluster fault-tolerant consistency control law are embedded into the motion control model of the system to achieve cross-domain collaboration and fault tolerance of heterogeneous unmanned systems under actuator failure. The effectiveness of the present invention is verified by simulation as follows:

[0236] In order to verify the effectiveness of the proposed control method, a heterogeneous unmanned system consisting of four unmanned boats and two fixed-wing UAVs was simulated and tested in MATLAB / SIMLINK. The structural parameters and aerodynamic parameters of the unmanned boats and UAVs are shown in Tables 1 and 2. Figure 3 The communication topology diagram of the above heterogeneous unmanned system is shown. It can be seen that the unmanned equipment in the system is divided into two clusters: unmanned boat group and drone swarms

[0237] Table 1 Structural parameters and aerodynamic parameters of unmanned boats

[0238]

[0239] Table 2 Structural parameters and aerodynamic parameters of fixed-wing UAV

[0240] parameter Value unit parameter Value unit <![CDATA[s i ]]> 1.463 <![CDATA[m 2 ]]> ρ 1.205 <![CDATA[kg·m -3 ]]> <![CDATA[m i ]]> 25 kg g 9.8 <![CDATA[m·s -2 ]]> <![CDATA[C iD0 ]]> 0.0225 - <![CDATA[C iDα ]]> 0.1002 <![CDATA[rad -1 ]]> <![CDATA[C iDα2 ]]> 1.0778 - <![CDATA[C iL0 ]]> 0.2153 - <![CDATA[C iLα ]]> 4.6333 <![CDATA[rad -1 ]]> <![CDATA[C iY0 ]]> 0 - <![CDATA[C iYβ ]]> -0.0046 <![CDATA[rad -1 ]]>

[0241] In the simulation, the output trajectories of the two cluster leaders are:

[0242]

[0243] The fault model of the heterogeneous unmanned system is selected as follows. The following fault occurs in unmanned boat 1 when t>5 seconds:

[0244]

[0245] When t>10 seconds, the following fault occurs in unmanned boat 2:

[0246]

[0247] When t>13 seconds, the following fault occurs in unmanned boat 3:

[0248]

[0249] When t>7 seconds, the following fault occurs in unmanned boat 4:

[0250]

[0251] Drone 1 has the following fault when t>10 seconds:

[0252]

[0253] Drone 2 has the following fault when t>8 seconds:

[0254]

[0255] The user-specified time-varying scalar functions β0(t) and β1(t) are chosen as follows:

[0256]

[0257] Among them, ζ0=0.15, ζ1=0.08,

[0258] The initial state of the heterogeneous unmanned system is shown in Table 3. The control parameters are selected as ι2=5,k 11 =k 21 =k 31 =k 41 =1.5, k 51 =k 61 =0.8, k 12 =k 22 =k 32 =k 42 =30, k 52 =k 62 =16, k i3 =1.12, τ i =0.05, T imax = 100, i = 1, 2, 3, 4, 5, 6. The number of basis functions is set to m = 11, and the desired cluster formation configuration is x 1d =[5;0;0],x 2d =[0;-5;0];x 3d =[-5;0;0];x 4d =[0;5;0];x 5d =[25;0;25];x 6d =[25; 0; -25]. In practice, the control input of the heterogeneous unmanned system is bounded, so in the simulation, the upper and lower limits of the unmanned boat control signal are set to u i0max =[500,500,500] T and ui0min =-[500,500,500] T , set the upper and lower limits of the drone control signal to u i0max =[1,1,1] T and u i0min =[0,-1,-1] T . Due to the UAV's sideslip angle β i Does not affect the control input u i0 , so the sideslip angle is not considered in the simulation.

[0259] Table 3 Initial state values ​​of heterogeneous unmanned systems

[0260]

[0261] Figure 4 Provides the output of the distributed reference state estimator Among them, unmanned boats 1-4 realize the output trajectory of the leader 1 of their cluster With accurate estimation, drone 1-2 achieves the output trajectory of the leader 2 in its cluster It can be seen that even in the case of failure of the actuators in the heterogeneous unmanned system, the output of the cluster leader can be accurately estimated by the proposed estimator, thus proving the effectiveness of the proposed estimator.

[0262] The temporal response of the cluster consensus error of heterogeneous unmanned systems is as follows: Figure 5 It can be observed that even in the case of executor failures, the cluster consensus error z of each system member is i0 You can also The time response of the position tracking error of the heterogeneous unmanned system is as follows: Figure 6 Similarly, the position tracking error z of each system member is i1 You can also The proposed control method converges to the preset accuracy limit. i0 (0) and z i1 (0) No excessive constraints are imposed, thus demonstrating the superiority of the proposed control method.

[0263] Figure 7 The lumped fault term h of the heterogeneous unmanned system is shown i and its estimated value It can be clearly seen that when the actuator fails, the neural network can approximate the lumped fault term very well, which shows that the new weight update law designed based on the immersion and invariant adaptation theory is effective for the neural network. Figure 8The control input of the heterogeneous unmanned system is provided. It can be observed that when the system encounters actuator failure, the control input of each system member can be quickly adjusted through the proposed control method to stabilize the formation in time.

[0264] In summary, the control method designed in the present invention can effectively achieve cluster consistency and fault tolerance of heterogeneous unmanned systems under actuator failures, and can ensure that the system error converges to a specified range within a preset time.

Claims

1. A cluster fault-tolerant consistency control method for heterogeneous unmanned systems, characterized in that: The method comprises the following steps: Step 1: For a heterogeneous unmanned system consisting of N1 unmanned boats and N2 fixed-wing UAVs, a collaborative motion control model of the heterogeneous unmanned system under actuator failure is established; Step 2: Design an error conversion mechanism to achieve performance constraint control of heterogeneous unmanned systems; Step 3, designing a distributed reference state estimator, wherein the distributed reference state estimator is used to estimate the output information of the cluster leader; Step 4. Under the framework of dynamic surface control, a cluster fault-tolerant consistency control scheme with preset time and predefined accuracy is designed based on the error conversion mechanism and neural network approximation strategy, as well as the output information of the cluster leader, so as to achieve cross-domain collaboration and fault tolerance of heterogeneous unmanned systems under actuator failure.

2. According to claim 1, a cluster fault-tolerant consistency control method for heterogeneous unmanned systems is characterized in that: The step 1 specifically includes the following process: Step 1.1: Establish the dynamic model of the i-th unmanned boat as follows: in, is the Jacobi transformation matrix, η i =[x i ,y i , ψ i ] T is the position and heading of the i-th unmanned boat in the inertial coordinate system, is η i The derivative of i =[u i , v i , r i ] T is the sway, drift and pitch speed of the i-th unmanned boat in the attached coordinate system, v i The derivative of i is the control input, M i is the inertia matrix, C i (v i ) is the Coriolis centripetal force matrix, D i (v i ) is the damping matrix; Define x i1 =η i , x i2 =v i ,u i =τ i , convert formula (1) into the following form: Among them, f i1 =J(ψ i )v i -v i and is a nonlinear term, g i1 =I3 and is the gain term, I3 represents the 3×3 unit matrix; Step 1.2: Establish the dynamic model of the i-th fixed-wing UAV as follows: Among them, x i ,y i 、z i is the position of the i-th fixed-wing UAV in the inertial coordinate system, and x i ,y i and z i The derivative of V i is the speed of the i-th fixed-wing UAV, V i The derivative of and are the heading angle and track angle of the i-th fixed-wing UAV, respectively. and They are respectively i and γ i The derivative of m i is the mass of the i-th fixed-wing UAV; g is the acceleration due to gravity; μ i , α i and β i are the inclination angle, attack angle and sideslip angle of the i-th fixed-wing UAV respectively; T i , D i , Y i and L i are respectively the thrust, drag, lateral force and lift exerted on the i-th fixed-wing UAV; The specific expression is as follows: Among them, C iL , C iD and C iY They are lift coefficient, drag coefficient and force coefficient respectively, and have the following forms: Among them, T imax Indicates the maximum thrust of the engine; For thrust throttle settings; represents dynamic pressure and ρ0 represents air density; s i is the wing area; C iL0 , C iLα , C iD0 , C iDα , C iDα2 , C iY0 , C iYβ is the aerodynamic coefficient; Define x i1 =[x i ,y i , z i ] T , x i2 =[V i , χ i , γ i ] T , Convert formula (3) into the following form: Among them, g i1 =I3 and is the gain term, f i1 =[f i11 , f i12 , f i13 ] T and f i2 =[f i21 , f i22 , f i23 ] T is a nonlinear term, expressed as follows: f i11 =V i cosγ i cosχ i -V i f i12 =V i cosγ i sinx i -x i f i13 =V i sing i -c i f i21 =-D i / m i -g sinγ i Step 1.3: Build actuator fault model: in i =ρ i in i0 +b i (5) Among them, ρ i =diag{ρ i1 , ρ i2 , ρ i3 } is the actuator efficiency loss factor, which is a diagonal matrix satisfying ρ i1 , ρ i2 , ρ i3 ∈(0,1]; b i is a bounded deviation fault vector; u i0 It is the command control input signal; Step 1.4: Substitute equation (5) into equation (2) and equation (4) to obtain the collaborative motion control model of heterogeneous unmanned systems under actuator failure: Among them, h i =g i2 [(ρ i -I3)u i0 +b i ] is the lumped fault item.

3. According to claim 1, a cluster fault-tolerant consistency control method for heterogeneous unmanned systems is characterized in that: The step 2 specifically includes the following process: The error conversion mechanism is as follows: Where, π(t) = 1-β(t) 2 z(t) 2 , z(t) represents the error variable of the system, β(t) is a time-varying scalar function, and the derivative of s(t) is: in, The derivative of the error conversion mechanism is transformed as follows: in, Omit (t).

4. According to claim 1, a cluster fault-tolerant consistency control method for heterogeneous unmanned systems is characterized in that: The step three specifically includes the following process: In order to estimate the output state of the cluster leader, the following distributed reference state estimator is established that can meet the preset performance requirements. First, given the,th cluster The leader system state equation is as follows: Among them, f l 0 and is a nonlinear function, is the system status, is the system output, is the control input. Secondly, the distributed reference state estimator is: Among them, there are: in, and l2 are control parameters, and Representation Matrix The maximum eigenvalue of Representation Matrix The minimum eigenvalue of ; is a 3×3 diagonal matrix that satisfies satisfy is the output of the distributed reference state estimator; express y i The estimate, y i is a constant and satisfies and y i Respectively represent y i The upper and lower bounds of i0 =[s i01 ,s i02 ,s i03 ] T is the conversion error variable obtained by changing the error conversion mechanism, and its expression is as follows: in, β0 is a time-varying scalar function defined by the user, z i0 =[z i01 , z i02 , z i03 ] T is the cluster consistency tracking error, which is defined as follows: in, is the output of the leader of the cluster where the ith system member is located. The subscript li indicates that the ith system member belongs to the ith cluster, that is, 5. According to the cluster fault-tolerant consistency control method for heterogeneous unmanned systems according to claim 1, it is characterized in that: The step 4 specifically includes the following process: Step 4.1: Design virtual control law definition is the position tracking error of the i-th system member, where x id represents the desired cluster formation configuration, and using the error conversion mechanism, we get: in, β1 is a user-defined time-varying scalar function, s i1 =[s i11 ,s i12 ,s i13 ] T ; Design a virtual control law for system (6) for: Among them, k i1 is a control parameter, is x id The derivative of is μ i11 The derivative of i11 =diag{μ i11q }, q = 1, 2, 3 is a 3 × 3 diagonal matrix, satisfying μ i12 =[μ i121 , μ i122 , μ i123 ] T satisfy Design the command filter as: Among them, τ i is the filter error, is the filter output, for The derivative of , the filtering error of equation (18) is defined as Step 4.2: Design the actual control law u i0 , definition is the speed tracking error of the i-th system member, then for system (6), the actual control law u i0 Designed for: Among them, k i2 To control the parameters, is the lumped fault term h i An estimated value of Step 4.3: Design the lumped fault term h i Estimated value of Lumped fault term h in the cooperative motion control model of heterogeneous unmanned systems under actuator failure i Using radial basis neural network approximation, it is expressed as follows: Among them, W i * represents the ideal weight vector, is the Gaussian basis function vector, is the approximate error, satisfying ε a is a constant matrix; Using immersion and invariant adaptation theory to obtain W i * The estimated value of , the approximate value of formula (18) is expressed as: in, W i * Asymptotic estimate of ; using θ i , Φ i and ε i express and Using the immersion and invariant adaptation theory, The update law and function θ i The design is as follows: in, Represents θ i right The partial derivative of k i3 is the control parameter. Let θ ipq Represents the matrix θ i The (p, q)th element in , we get: Among them, Φ ip Represents Φ l The pth element in .

Images