A dynamic event-triggered control method for surface ships based on composite learning
Through dynamic event-triggered adaptive neural output feedback control method based on composite learning, the problem of composite uncertainty reconstruction under unknown surface ship speed is solved, the control performance and accuracy is improved, mechanical wear is reduced, and disturbance estimation technology is expanded. Application scope.
Patent Information
- Application Number
- CN202211125909.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-16
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-09-16
AI Technical Summary
In the motion control of surface ships, when the ship speed is unknown, it is difficult for the prior art to effectively reconstruct internal and external composite uncertainty, and event trigger control methods are less used in this case, resulting in control performance and mechanical wear problems .
Adopting adaptive neural output feedback control method based on composite learning, by establishing a ship mathematical model, designing neural network state observers and control laws, introducing serial parallel estimation models, embedding prediction errors and composite perturbations estimation values, and designing composites Learn the law and establish a dynamic event trigger protocol between the control law and the actuator.
The composite uncertainty is rapidly and accurate online reconstruction under unknown ship speed, reducing the mechanical wear of the actuator, improving control performance and accuracy, and expanding the application range of disturbance estimation technology.
Smart Images

Figure CN115562005B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of automatic control of surface ships, and in particular relates to a surface ship dynamic event triggering control method based on composite learning. Background Art
[0002] Surface ships are essential for various marine operations, such as marine exploration, marine transportation, and marine survey. During the navigation process, the navigation environment of ships is complex and changeable, and they are often affected by wind, waves, and currents. At the same time, there are many uncertainties in the ship itself, such as uncertainty in ship parameters, measurement uncertainty in sensor equipment, sudden failure of propulsion system, etc., which bring great challenges to the safe navigation of intelligent ships. Due to the external environment and ship maneuvering characteristics, the dynamic part of the ship is inevitably affected by internal dynamic uncertainty and unknown external disturbances. It is worth noting that dealing with these uncertainties in motion control design will have a significant impact on control performance, such as control accuracy, stabilization time, and stability performance. In response to the above problems, the relevant prior art is disclosed as follows: In order to cope with the adverse effects caused by internal and / or external uncertainties, the literature "Disturbance observer-based adaptive intelligent control of marine vessel with position and heading constraint condition related to desired output" and the literature "Event-triggered finite-time tracking control of underactuated MSVs based on neural network disturbance observer" use disturbance observers based on adaptive neural networks to reconstruct composite uncertainties including internal dynamic uncertainties and unknown external disturbances. In order to further improve the reconstruction accuracy of composite uncertainties, the paper "Composite neural dynamic surface control of a class of uncertain nonlinear systems in strict-feedback form" introduces a serial-parallel estimation model (SPEM), constructs the same parallel dynamics as in the paper "Composite neural learning fault-tolerant control for underactuated vehicles with event-triggered input", and designs an improved adaptive law. It should be pointed out that the above work generally requires the ship speed to be known. Obviously, due to sensor failure or other problems, the ship's speed information may not be directly available.To address this problem, the paper "Event-triggered composite adaptive fuzzy output-feedback control for path followingof autonomous surface vessels" designed a fuzzy-based state observer based on the idea of a neural network state observer, introduced SPEM, and then developed an adaptive fuzzy output feedback path tracking control scheme based on composite learning. It should be pointed out that although the above work effectively solves the problem of unknown ship speed information, it does not fully consider the problem of disturbance estimation accuracy. To address this problem, in the existing work, there are few papers on designing disturbance observers under conditions of unknown ship speed. Whether it is possible to classify and reconstruct internal and external uncertainties in the absence of ship speed information has always been an unresolved issue.
[0003] From a practical point of view, the actuators installed on the ship are subject to physical constraints such as response frequency. In addition, mechanical wear will be aggravated due to vibration of the hull and equipment, measurement noise, etc. In order to reduce mechanical wear and ensure that the control command remains within the working range of the actuator, an effective measure is to reduce the transmission frequency of the control command between the control law and the actuator. To achieve this, the document "Event-triggered approximate optimal path-following control for unmanned surface vehicles with state constraints" adopts the event-triggered control (ETC) method to solve the path following problem. Compared with continuous-time control, the control scheme under ETC not only reduces the number of transmissions of control commands calculated by the control law, but also maintains satisfactory control performance. It should be pointed out that the event-triggered protocol (ETP) includes internal triggering, external triggering and compound triggering. The ETP proposed in the above literature is a static ETP. Compared with the static ETP, the dynamic triggering protocol (DETP) can further reduce the number of events while maintaining acceptable control performance. However, the application of the dynamic triggering control method to solve the control problem with internal and external uncertainties when the ship speed is unknown is rarely discussed in the existing literature.
[0004] Therefore, it is urgent to develop a dynamic event triggered control method for surface ships based on composite learning to solve the above problems. Summary of the invention
[0005] The purpose of the present invention is to provide a surface ship dynamic event triggered adaptive neural output feedback control method based on composite learning.
[0006] In order to solve the above technical problems, the present invention provides a surface ship dynamic event triggering control method based on composite learning, comprising:
[0007] Step 1: Establish a mathematical model of the ship;
[0008] Step 2: Design a neural network state observer;
[0009] Step 3: Design the control instructions of the control law.
[0010] The present invention makes full use of the unique advantages of adaptive neural networks and disturbance observers to solve the problem of reconstructing and compensating for internal and external uncertainties when the ship speed is unknown. In order to improve the control performance, a serial-parallel estimation model is introduced under the backstepping design framework to improve the reconstruction accuracy, and a composite learning law is designed by embedding the estimated values of the prediction error and the composite disturbance. A dynamic event triggering protocol is established between the control law and the actuator to reduce the mechanical wear of the actuator.
[0011] Further,
[0012] The ship mathematical model established in step 1 is:
[0013]
[0014]
[0015] Where η=[x,y,ψ] T Represents the position (x, y) and heading angle ψ of the ship in the inertial coordinate system;
[0016] υ=[u,v,r] T is the velocity vector of the ship in the appendage coordinate system, and u, v, and r represent the forward velocity, transverse drift velocity, and bow angular velocity of the ship in the hull coordinate system, respectively;
[0017] R(ψ) is given by The given rotation matrix;
[0018] M∈R 3×3 is the inertia matrix;
[0019] f(υ) is the nonlinear dynamics, including the Coriolis centripetal term and the damping term;
[0020] τ=[τ1,τ2,τ3] T is the control input vector generated by the propulsion system;
[0021] τ d =[τ d,1 ,τ d,2 ,τ d,3 ] TIt is the unknown, time-varying external disturbance vector caused by wind, current, waves, etc.
[0022] Further,
[0023] The step 2 includes: based on the ship mathematical model, using adaptive neural network and disturbance observer technology, a classification reconstruction idea is proposed, and a new adaptive neural network state observer is designed through the coupling design between the state observer and the disturbance observer. It is used to reconstruct the composite uncertain influence suffered by the surface ship when the speed is unknown:
[0024] According to equations (1) and (2), the dynamics of the ship can be expressed as
[0025]
[0026] Where β(υ,τ)=Μ -1 (τ-f(υ)), Considering Assumption 2, β(υ,τ) and It cannot be used in the design of neural network-based state observers. According to Lemma 3, the unknown term β(υ,τ) can be reconstructed online by neural network approximation techniques;
[0027]
[0028] In the formula, ζ=[υ T ,τ T ] T , is the weight matrix, is the basis function vector, and ε o is the approximation error vector. Here ρ i (ζ)=[ρ i,1 (ζ),…,ρ i,l (ζ)] T And ||ε o ||≤ε ι where ε ι >0 is an unknown constant;
[0029] further, Can be written as
[0030] In the formula, is a composite perturbation. According to the approximation theory of neural networks and assumption 1, there is an unknown constant satisfy in,
[0031] Assumption 1: The external disturbance τ d Its derivative is unknown and bounded, that is, there is an unknown constant and satisfy and
[0032] The position, velocity, compound disturbance and The estimate is defined as and Right now and and Obtained by the following state observer:
[0033]
[0034] In the formula, σ is an auxiliary variable, κ>0, κ1>0 and κ2>0 are observer gains;
[0035] To obtain an estimate of the composite disturbance ω Design the following auxiliary variable δ o ∈R 3 :
[0036] δ o =ω-κ o υ; (7)
[0037] In the formula, κ0>0 is a design constant. From formula (7), we can see that if and Then we can get the estimate of the composite disturbance ω, and according to formula (6), we can get
[0038] To obtain Design the following dynamics:
[0039]
[0040] Furthermore, the estimated value of the composite disturbance is:
[0041]
[0042] remember Can get
[0043] Further, take The derivative of , and using equations (5) and (7)-(9) we can get
[0044]
[0045] In the formula,
[0046] The error of the neural network state observer is:
[0047]
[0048] In the neural network state observation, the adaptive law of the neural network is designed as:
[0049]
[0050] In the formula, K o =diag(K o,1 ,K o,2 ,K o,3 ) is a design matrix, K o,i is a constant, k3 is a design constant.
[0051] Further,
[0052] The step 3 includes: obtaining the prediction error by introducing the series-parallel estimation model according to the mathematical model of the surface ship, then designing the composite learning law by embedding the prediction error and the composite disturbance estimation, and establishing a dynamic event triggering protocol between the control law and the actuator:
[0053] Define the position error e η ∈R 3 and speed error e υ ∈R 3 ;
[0054] e η =η-η r ; (13)
[0055] e υ =υ-α; (14)
[0056] In the formula, α∈R 3 is α υ ∈R 3 The filtering form, α∈R 3 can be obtained by the following multivariate filter
[0057]
[0058] Where J>0 is the time constant of the filter; get
[0059] For η Differentiate and substitute into the above formula to get
[0060]
[0061] Design virtual control function α υ
[0062]
[0063] Where c1∈R 3×3 is a symmetric positive definite design matrix;
[0064]
[0065] For υ Differentiate and substitute into formula (2) to obtain
[0066]
[0067] In the formula, and According to Assumptions 1 and 2, there are two unknown terms in Equation (19), namely, F(υ) and τ d . F(υ) and τ are not used directly in the control law τ d According to the online reconstruction theory of neural networks, F(υ) can be reconstructed online, but τ d It cannot be reconstructed online because its states η and υ are independent;
[0068] Among them, assumption 2 is: the inertia matrix M includes the standard part M0 and the uncertain part δM, that is, M = M0 + δM, and δM and f(υ) are completely unknown;
[0069] Further, using Lemma 2, we can get
[0070]
[0071] In the formula, ρ(ν)=[ρ1(ν) T ,ρ2(ν) T ,ρ3(ν) T ] T , ε υ =[ε υ,1 ,ε υ,2 ,ε υ,3 ] T .
[0072] According to the NN approximation principle, the error vector ε υ satisfy and is a constant;
[0073] remember According to hypothesis 1 and There are unknown constants Make In addition, formula (20) and Substituting into formula (19) we get
[0074]
[0075] The trajectory tracking control law of the surface ship is designed as follows:
[0076]
[0077]
[0078] Where c2∈R 3×3 is a symmetric positive definite design matrix, yes The estimate, is τ w An estimated value of
[0079] The dynamic event triggering protocol is:
[0080]
[0081] a ι and b ι is the design constant, θ ι By the following dynamic
[0082]
[0083] In the formula, ρ ι ∈R + is the design constant, θ ι (0) is θ ι The initial value of .
[0084] Estimation of composite disturbances It can be obtained by the following disturbance observer;
[0085]
[0086] In the formula, is the auxiliary variable δ w Here, where κ w is a user-designed constant. In this work, to obtain δ w The following dynamics are designed:
[0087]
[0088] In the formula, is a positive definite design matrix, μ∈R 3×3 is a prediction error variable designed in the follow-up process;
[0089] The prediction error variable is defined as
[0090]
[0091] In the formula, Yes υ The predicted value of
[0092] Obtaining predictions by building a serial-parallel estimation model
[0093]
[0094] Adaptive Law Designed for:
[0095]
[0096] In the formula, K c ∈R 3×3 is a positive definite design matrix, κ c ∈R + is the design constant.
[0097] Compared with the prior art, the present invention has the following beneficial effects:
[0098] The present invention reconstructs the nonlinear dynamic uncertainties in the longitudinal and steering in the ship mathematical model by adopting neural network approximation technology, establishes a finite-time disturbance estimator, and reconstructs the overall uncertainty in the longitudinal and steering including unknown external disturbances and inaccessible parts online. It can realize fast and accurate online reconstruction of complex uncertainties, release the precise requirements of online interference estimation technology on the ship motion model, expand the application scope of online interference estimation technology, introduce event-triggered control method, and suppress unnecessary wear of the drive.
[0099] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practicing the present invention.
[0100] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0101] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0102] Figure 1 It is a flow chart of the present invention;
[0103] Figure 2A graph of the tracking performance of the invention in the xy plane;
[0104] Figure 3 The reference position and actual position response curve diagram provided by the present invention;
[0105] Figure 4 Tracking error variation curve diagram of the present invention;
[0106] Figure 5 The control input curve diagram of the present invention;
[0107] Figure 6 A curve diagram showing a change in the norm of the weight vector estimate value of the present invention;
[0108] Figure 7 The speed and estimated value thereof of the present invention;
[0109] Figure 8 A graph showing the variation of the observer error of the present invention;
[0110] Fig. 9 The trigger time and the response diagram of the trigger moment of the present invention. DETAILED DESCRIPTION
[0111] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0112] Example 1
[0113] In this embodiment, if Figures 1 to 9 As shown, this embodiment provides a surface ship dynamic event triggering control method based on composite learning, and the ship dynamic event triggering control method includes:
[0114] Step 1: Establish a mathematical model of the ship;
[0115] Step 2: Design a neural network state observer;
[0116] Step 3: Design the control instructions of the control law.
[0117] The present invention makes full use of the unique advantages of adaptive neural networks and disturbance observers, proposes a classification reconstruction idea, and realizes the design of a coupler between the state observer and the disturbance observer. It solves the control problem under the influence of internal and external uncertainties when the ship speed is unknown. Furthermore, in order to improve the control performance, under the backstepping design framework, in order to improve the reconstruction accuracy, a serial-parallel estimation model is introduced, and a composite learning law is designed by embedding the estimated values of the prediction error and the composite disturbance. A dynamic event triggering protocol is established between the control law and the actuator to reduce the mechanical wear of the actuator. Finally, a new dynamic event-triggered composite learning adaptive neural output feedback control scheme is proposed. Compared with previous studies, the control scheme designed by the present invention can realize the rapid and accurate online reconstruction of composite uncertainties, release the precise requirements of the online disturbance estimation technology for the ship motion model, expand the application scope of the online disturbance estimation technology, introduce the dynamic event triggering control method, and suppress the unnecessary mechanical wear of the actuator.
[0118] In this embodiment, the method for establishing a ship mathematical model includes:
[0119]
[0120]
[0121] Where η=[x,y,ψ] T Represents the position (x, y) and heading angle ψ of the ship in the inertial coordinate system;
[0122] υ=[u,v,r] T is the velocity vector of the ship in the appendage coordinate system, and u, v, and r represent the forward velocity, transverse drift velocity, and bow angular velocity of the ship in the hull coordinate system, respectively;
[0123] R(ψ) is given by The given rotation matrix;
[0124] The rotation matrix R(ψ) has the following properties:
[0125] Property 1: The matrix R(ψ) is orthogonal, that is, ||R(ψ)||=1 and R -1 (ψ) = R T (ψ);
[0126] Property 2: in
[0127] Note 1: According to property 2, we can get Right now
[0128] M∈R 3×3is the inertia matrix;
[0129] f(υ) is the nonlinear dynamics, including the Coriolis centripetal term and the damping term;
[0130] τ=[τ1,τ2,τ3] T is the control input vector generated by the propulsion system;
[0131] τ d =[τ d,1 ,τ d,2 ,τ d,3 ] T is the unknown, time-varying external disturbance vector caused by wind, current, waves, etc.;
[0132] To facilitate design and analysis, the following assumptions are made:
[0133] Assumption 1: External disturbance τ d Its derivative is unknown and bounded, that is, there is an unknown constant and satisfy and
[0134] Assumption 2: The inertia matrix M includes the standard part M0 and the uncertain part δM, that is, M = M0 + δM, and δM and f(υ) are completely unknown;
[0135] Assumption 3: Reference trajectory η r =[x r ,y r ,ψ r ] T It has first and second order derivatives;
[0136] Assumption 4: The speed υ is unknown in the control design;
[0137] For the design and analysis of ship tracking control, the following definitions and theorems are given:
[0138] Lemma 1: For any scalars p and q and positive constants b, m and n, if m and n satisfy m>1, n>1 and (m-1)(n-1)=1, we have
[0139]
[0140] Lemma 2: For any given nonlinear function R n →R is defined on a compact set There is a NN So for any given positive scalar ε,
[0141]
[0142] In the formula, and are weight vector and basis function vector respectively, satisfying and here, and ε are constants, l>1 is the number of nodes. Basis function Choose a Gaussian function, that is:
[0143]
[0144] In the formula, ι′ i =[ι i,1 ,…,ι i,l ] T and ω i for Center point vector value and width.
[0145] Lemma 3: For a radial basis function vector with a Gaussian function, if is the input vector, and and are constants and bounded vectors respectively, there exists a bounded function vector Make
[0146]
[0147] In the formula, is a constant.
[0148] Design of state observer based on neural network
[0149] According to equations (1) and (2), the dynamics of the ship can be written as
[0150]
[0151] Where β(υ,τ)=Μ -1 (τ-f(υ)),
[0152] Considering Assumption 2, β(υ,τ) and It cannot be used in the design of neural network based state observer.
[0153] According to Lemma 3, the unknown term β(υ,τ) can be reconstructed online by neural network approximation technique.
[0154]
[0155] In the formula, ζ=[υ T ,τ T ] T , is the weight matrix, is the basis function vector, and ε o is the approximation error vector. Here ρ i (ζ)=[ρ i,1 (ζ),…,ρ i,l (ζ)] T And ||ε o ||≤ε ι where ε ι >0 is an unknown constant.
[0156] further, Can be written as
[0157]
[0158] In the formula, is the lumped perturbation. According to the approximation theory of neural networks and assumption 1, there is an unknown constant satisfy
[0159] Position, velocity, concentrated disturbance and The estimate is defined as and Right now and and is obtained by the following neural-based state observer:
[0160]
[0161] In the formula, σ is an auxiliary variable, κ>0, κ1>0 and κ2>0 are observer gains.
[0162] To obtain an estimate of the concentrated disturbance ω Design the following auxiliary variable δ o ∈R 3 :
[0163] δ o =ω-κ o υ; (12)
[0164] In the formula, κ0>0 is a design constant. From formula (12), we can see that if we obtain and Then we can get the estimate of the composite disturbance ω, and according to formula (11), we can get
[0165] To obtain Design the following dynamics:
[0166]
[0167] Further, the composite disturbance is estimated as:
[0168]
[0169] remember Can get
[0170] Further, take The derivative of , and using equations (10) and (12)-(14) we can get
[0171]
[0172] In the formula,
[0173] The error of the neural network state observer is:
[0174]
[0175]
[0176] In the neural network state observation, the weight adaptation law of the neural network is designed as:
[0177]
[0178] In the formula, K o =diag(K o,1 ,K o,2 ,K o,3 ) is a design matrix, K o,i is a constant, k3 is a design constant.
[0179] Control law design
[0180] For MSVs with internal / external uncertainties and unmeasurable speed, an event-triggered adaptive neural output feedback control law is designed in the backstepping design framework using methods such as neural-based state observer (10), adaptive neural technology, and disturbance observer technology. The entire design process includes two steps.
[0181] Before control design, define the following position error e η ∈R 3 and speed error e υ ∈R 3
[0182] e η =η-η r ; (18)
[0183] eυ =υ-α; (19)
[0184] In the formula, α∈R 3 is α υ ∈R 3 The filtering form, here, α∈R 3 can be obtained by the following multivariate filter
[0185]
[0186] Where J>0 is the time constant of the filter. get
[0187] For η Differentiating and using (1), (28) and (29) yields
[0188]
[0189] Design virtual control function α υ
[0190]
[0191] Where c1∈R 3×3 is a symmetric positive definite design matrix.
[0192] Using formula (22), we can get
[0193]
[0194] Step 2: Differentiate υ And using (2) we get
[0195]
[0196] In the formula, and According to assumptions 1 and 2, we know that there are two unknown terms in (24), namely F(υ) and τ d Obviously, F(υ) and τ are not used directly in the control law τ. d According to the online reconstruction theory of NN, F(υ) can be reconstructed online, but τ d It cannot be reconstructed online because its states η and υ are independent.
[0197] Further, using Lemma 2, we can get
[0198]
[0199] In the formula, ρ(ν)=[ρ1(ν) T ,ρ2(ν)T ,ρ3(ν) T ] T , ε υ =[ε υ,1 ,ε υ,2 ,ε υ,3 ] T According to the NN approximation principle, the error vector ε υ satisfy and is a constant.
[0200] remember According to hypothesis 1 and There are unknown constants Make In addition, (25) and Substituting (24) into (24) yields
[0201]
[0202] In addition, the trajectory tracking control law of the ship is designed as follows
[0203]
[0204]
[0205] Where c2∈R 3×3 is a symmetric positive definite design matrix, yes The estimate, is τ w estimates;
[0206] Dynamic Event Triggered Protocol (DETP)
[0207]
[0208] a ι and b ι is the design constant, θ ι By the following dynamic
[0209]
[0210] In the formula, ρ ι ∈R + is the design constant, θ ι (0) is θ ι The initial value of .
[0211] Estimation of composite disturbances It can be obtained by the following disturbance observer
[0212]
[0213] In the formula, is the auxiliary variable δ w Here, where κ w is a user-designed constant. In this work, to obtain δ w The following dynamic
[0214]
[0215] In the formula, is a positive definite design matrix, μ∈R 3×3 is a prediction error variable designed in the subsequent process.
[0216] The prediction error variable is defined as
[0217]
[0218] In the formula, Yes υ The predicted value of .
[0219] Obtaining predictions by building a serial-parallel estimation model
[0220]
[0221] Adaptive Law Designed for:
[0222]
[0223] In the formula, K c ∈R 3×3 is a positive definite design matrix, κ c ∈R + is the design constant.
[0224] A simulation verification study was conducted on the dynamic event-triggered adaptive neural control method for surface ships based on composite learning to verify its effectiveness.
[0225] During the simulation test, the model parameters are as follows:
[0226] m 11 =200, m 22 =250, m 33 =80,d 11 =70,d 22 =100,d 23 =40,d 32 =40,d 33 =80.
[0227] Reference trajectory η r Depend on Equation generated.
[0228] In the formula, τ r =[1,0.2cos 2 (0.01πt),0.3sin 2 (0.01πt)] T .
[0229] The external disturbance is set as: here, By a first-order Markov process generate, is a zero-mean Gaussian white noise process. w It is generated by the superposition of the following time-varying functions:
[0230]
[0231] The design parameters are selected as κ=120, κ1=4.5, κ2=10, κ o =1,κ3=5,κ c =0.01, J=0.01, K o =diag([30I7,30I7,30I7])、c1=diag([0.6,0.5,0.7])、ρ ι =(1,1,1),κ w =diag([6,6,3]), κ p =diag([10,10,5])、Π=diag(2,2,3)Φ=diag(3,3,2)、c2=diag([35,35,42])、a ι =(2,1,1),b ι =(2,1.5,0.5).
[0232] The initial conditions are chosen as: υ(0)=[0.1m / s,0.1m / s,0.01rad / s] T , and the rest are set to 0.
[0233] β(ζ) and The number of nodes are 6 and 12 respectively, and the centers are [-2,2]×[-2,2]×[-2,2]×[-20,20]×[-25,25]×[-20,20] and Evenly distributed within the range, the width is set to ω l,o =4 and ω l,υ =2.
[0234] This embodiment uses software to conduct computer simulation research, and the results are as follows Figures 2 to 9 In addition, in order to prove the superiority of the proposed control scheme in work, the simulation is compared with the continuous-time adaptive neural composite learning control scheme proposed in the literature "Control of fully actuated ocean surface vessels using a class of feedforward approximators" and the adaptive neural output feedback control scheme based on high gain observer. Figure 2-3 The tracking control performance of the ship is shown in Figure 2. It can be found that the proposed scheme can force the ship to follow the reference trajectory and obtain satisfactory control performance. The curves of position error and heading error are shown in Figure 2. Figure 4 As shown, this indicates that the position error is bounded and the proposed control scheme can guarantee satisfactory control accuracy. Meanwhile, the adaptive neural network and disturbance observer can effectively compensate for the compound uncertainties. Figure 5 The variation curve of the control input is plotted, from which it can be clearly seen that the actuator control input is bounded and reasonable. Figure 6 The 2-norm curve of the neural network weights is shown, which means it is bounded. Figure 7 The curve showing the change of ship speed and its estimated value, Figure 8 The estimated error curves of the ship's position and heading are plotted. Figure 7-8 It is shown that the neural network-based state observer can accurately reconstruct the ship's speed when the ship is subject to internal and external uncertainties. Fig. 9 The instant of event triggering and the time interval between event triggering are shown, which clearly shows that the actuator control command is not transmitted indefinitely, that is, the Zeno phenomenon caused by the ETC method does not occur. Figure 2-3 , we can clearly find that continuous-time control and control based on high-gain observer can also force the ship to follow the predetermined reference trajectory, which is the same as the control scheme proposed in this invention. Figure 4 It can be clearly seen from Table 1 that the tracking control accuracy of the ETC scheme proposed in the present invention is almost the same as that of the continuous control scheme, and is better than the tracking control accuracy of the control scheme based on the high-gain observer. This is because the adaptive neural network cannot reconstruct the external disturbance. Figure 5 From Table 2, it can be concluded that the control inputs under the three control schemes are bounded and reasonable, but the average energy consumption of our proposed scheme is the highest, which means that ETC needs to compensate for the control command signal loss caused by ETC by increasing energy consumption.
[0235]
[0236] Table 1
[0237] In summary, the present invention considers the influence of internal and external composite disturbances on ships, and designs a dynamic event triggering control method based on composite learning. Taking full advantage of the special advantages of neural networks and disturbance estimation technology, a classification reconstruction idea is proposed, and it is applied to the control design of state observers and control laws. In the design of the state observer, a coupling design method of state observer and disturbance observer based on adaptive neural network is proposed. In order to improve the control performance, an improved adaptive law is designed by introducing a serial-parallel estimation model in the control law design. In addition, in order to reduce the mechanical wear of the actuator, a new dynamic event triggering protocol is developed by introducing additional dynamic variables. Finally, under the framework of inverse design, a new dynamic event triggering control scheme based on composite learning is proposed.
[0238] The various devices selected in this application (components whose specific structures are not described) are all universal standard parts or components known to technical personnel in this field, and their structures and principles can be known to technical personnel through technical manuals or through conventional experimental methods.
[0239] In the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0240] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.
[0241] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.
Claims
1. A surface ship dynamic event triggering control method based on composite learning, characterized in that: include: Step 1: Establish a ship mathematical model. The ship mathematical model is: Where η=[x,y,ψ] T Represents the position (x, y) and heading angle ψ of the ship in the inertial coordinate system; υ=[u,v,r] Τ is the velocity vector of the ship in the appendage coordinate system, and u, v, and r represent the forward velocity, transverse drift velocity, and bow angular velocity of the ship in the hull coordinate system, respectively; R(ψ) is given by The given rotation matrix; M∈R 3×3 is the inertia matrix; f(υ) is the nonlinear dynamics, including the Coriolis centripetal term and the damping term; τ=[τ1,τ2,τ3] T is the control input vector generated by the propulsion system; τ d =[τ d,1 ,τ d,2 ,τ d,3 ] T is the external disturbance vector; Step 2: Design a neural network state observer, including: According to the ship mathematical model, the following state observer is designed: In the formula, σ is an auxiliary variable, κ>0, κ1>0 and κ2>0 are observer gains; To obtain the auxiliary variable δ o ∈R 3 Estimated value of Design the following dynamics: Where κ0>0 is a design constant. and Then we can get the estimate of the composite disturbance ω; The estimated value of the composite perturbation is: remember Can get In the neural network state observation, the adaptive law of the neural network is designed as: In the formula, K o =diag(K o,1 ,K o,2 ,K o,3 ) is a design matrix, K o,i is a constant, k3 is a design constant; Step 3: Design the control instructions of the control law, including: designing a disturbance observer based on the mathematical model of the surface ship under the design framework of vector backstepping, and introducing the series-parallel estimation model to establish the predictor of the kinematic channel, so as to obtain the prediction error of the speed tracking error, and then using the prediction error, speed tracking error and composite disturbance estimation as the excitation conditions for parameter adaptation, so as to design the composite learning law; further, establishing a dynamic event triggering protocol between the control law and the actuator; specifically including: The trajectory tracking control law of the surface ship is designed as follows: Where c2∈R 3×3 is a symmetric positive definite design matrix, yes The estimate, is τ w An estimated value of The dynamic event triggering protocol in the control law is: a ι and b ι is the design constant, θ ι By the following dynamic In the formula, ρ ι ∈R + is the design constant, θ ι (0) is θ ι The initial value of The disturbance observer in the control law is: In the formula, is the auxiliary variable δ w Here, where κ w is a user-designed constant. w The following dynamics are designed: In the formula, is a positive definite design matrix, μ∈R 3×3 is a prediction error variable; Serial-parallel estimation model in control law to obtain prediction value In the formula, Yes υ The predicted value of Adaptive Law Designed for: In the formula, K c ∈R 3×3 is a positive definite design matrix, κ c ∈R + is the design constant.
