Underwater vehicle model reference adaptive control system and method based on transient performance optimization

Abstract

The invention provides an underwater vehicle model reference adaptive control system and method based on transient performance optimization. The system comprises a vehicle, a sensor, an execution mechanism and a controller. The control method comprises the following steps: aiming at a controlled object underwater vehicle, designing a model reference adaptive controller, and carrying out transient performance optimization on a control quantity. Wherein the control quantity u is composed of three parts, namely the system control quantity u = un + ua + v, un is the system nominal control quantity, ua is the additional control quantity designed by the neural network, and v is the transient performance optimization control quantity. By designing the control input u, the matching uncertainty of the system can be inhibited. The method has the advantages that the rapid stability-free control of the underwater vehicle is realized, the contradiction between the rapidity and the error of the control system is relieved, and the stability-free control effect is achieved; the engineering design cost can be reduced, and the rapid and stable self-adaptive control effect can be achieved under the conditions that a hydrodynamic test is not conducted and parameter errors are large.

Description

Technical Field

[0001] The present application belongs to the technical field of underwater vehicles, and specifically relates to an underwater vehicle model reference adaptive control system and method based on transient performance optimization. Background Art

[0002] The control technology of underwater vehicles refers to the underwater vehicle, in the underwater environment, according to its own movement, under the action of certain control laws, controlling the movement of each rudder or propeller, automatically sailing along a predetermined course in a specified posture. The control system controls the motion state of the underwater vehicle under the disturbance of the complex marine environment. Therefore, the control system of the vehicle is essentially a study of the control method of the controlled object under the action of random disturbance.

[0003] In actual navigation, underwater vehicles still have the following problems:

[0004] Underwater vehicles are complex multi-input multi-output (MIMO) systems with strong coupling and high nonlinearity. The traditional methods for designing their controllers are local linearization and PID control. However, traditional control methods require a relatively accurate understanding of the system's fluid dynamic model parameters, and the effectiveness of the designed controller is limited to the vicinity of the equilibrium point. The control system must meet the needs of large-scale maneuvers of the vehicle. When the vehicle makes large maneuvers, nonlinear model errors will lead to a decrease in control effect and reduced stability of the control system.

[0005] During navigation, the vehicle is subject to random interference from waves and currents. Traditional control methods will result in a slow system response speed and cannot meet the vehicle's requirements for fast response, strong robustness and high adaptability.

[0006] During navigation, actuator uncertainties such as rudder zero-point deviation, rudder force coefficient uncertainty and other unmodeled errors often affect the transient performance of the control system. Summary of the invention

[0007] The purpose of the present application is to overcome the defects of existing underwater vehicles, such as poor stability, slow response speed and low adaptability when some fluid parameters are unknown.

[0008] In order to achieve the above objectives, the present application proposes an underwater vehicle model reference adaptive control system based on transient performance optimization, the system comprising:

[0009] The system model of the system is:

[0010]

[0011] where x = [ψω] Tis the system state variable, ψ is the heading angle of the underwater vehicle, ω is the heading angular velocity of the underwater vehicle; T is the vector transpose; is the derivative of the system state variable, u(t) is the system input signal, that is, the rudder angle of the actuator; A and B are controllable matrices; d(x,t) is the uncertain term; t is time; Λ>0 is a diagonal matrix that satisfies the Lipschitz condition:

[0012]

[0013] Where l is a constant; ||·|| is the norm calculation;

[0014] The reference model of the system is:

[0015]

[0016] Among them, x r is the system state variable of the reference model, is the derivative of the system state variable of the reference model, y r Set the value for the output;

[0017] is the Hurwitz matrix with characteristic roots in the right half plane; K 1 To adjust the closed loop pole;

[0018] B r =BK 2 ; Parameter K 2 The selection principle is based on y r Adjust to ensure that the closed-loop system steady-state output is y r ; The system control quantity u of the system is expressed as:

[0019] u=u+u+v

[0020] Among them, u n is the nominal control quantity; u a is the additional control quantity; v is the transient performance enhancement control quantity.

[0021] As an improvement of the above system, the nominal control amount u n for:

[0022] u n =-K 1 x+K 2 y r .

[0023] As an improvement of the above system, the additional control amount u a for:

[0024]

[0025] Where Θ(x)=[Θ 0 (x),u n ] T , Θ 0 (x) is the basis function;

[0026] is the estimated weight.

[0027] As an improvement to the above system, the basis function θ 0 (x) is a set of Gaussian functions.

[0028] As an improvement of the above system, the transient performance enhancement control amount v is:

[0029]

[0030] in, μ is the filter time constant; s is the complex variable in Laplace transform.

[0031] As an improvement to the above system, the parameter K 2 The selection satisfies:

[0032] -CA -1 BK 2 =-I

[0033] Where I is the identity matrix; C is a matrix that satisfies:

[0034]

[0035] Among them, y is the system output value.

[0036] The present application also provides an underwater vehicle model reference adaptive control method based on transient performance optimization, which is implemented based on the above system and includes:

[0037] Build system models and reference models;

[0038] Determine the basis function according to the factors affecting the underwater vehicle track, calculate the nominal control quantity, additional control quantity and transient performance enhancement control quantity, and determine the adaptive control system;

[0039] By inputting the actuator rudder angle into the system model, the navigation attitude of the underwater vehicle can be output, completing the model reference adaptive control of the underwater vehicle.

[0040] Compared with the prior art, the advantages of this application are:

[0041] 1. The present invention can realize rapid and stable error-free control of underwater vehicles, alleviate the contradiction between the rapidity and error of the control system, and achieve a stable error-free control effect;

[0042] 2. The present invention can reduce the engineering design cost. When the hydrodynamic test is not performed and the parameter error is large, the actuator uncertainty is reduced. When the vehicle encounters a large disturbance, the steering gear zero point, the rudder force coefficient, and other unmodeled errors can be adaptively compensated to achieve a fast and stable adaptive control effect.

[0043] 3. The adaptive method can offset the influence of parameter uncertainty, but the adaptive process is the change process of the estimated parameters, which affects the transient response of the system and causes oscillation of the transient response. The method proposed in the present invention can effectively suppress the transient response oscillation phenomenon in the model reference adaptive control process and ensure the optimization of transient performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 Shown is the structural diagram of the model reference adaptive control system;

[0045] Figure 2 The figure shows a schematic diagram of the tracking signal; the green straight line is the step signal setting value of 5 meters; the blue dotted line is a schematic diagram of the control result of tracking the 5-meter setting value; the blue straight line is the continuous signal generated by the reference model; the purple dotted line is a schematic diagram of the control result of tracking the continuous signal;

[0046] Figure 3 Shown is the structure diagram of the model reference adaptive controller;

[0047] Figure 4 The figure shows the structure diagram of the transient performance enhancement model reference adaptive controller;

[0048] Figure 5 Shown are the simulation results of model reference adaptive heading control;

[0049] Figure 6 Shown are simulation results for model-referenced adaptive heading control with transient performance enhancement. DETAILED DESCRIPTION

[0050] The technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0051] In order to solve the control problem of underwater vehicles, nonlinear models are needed to describe their complex underwater motion process. In order to solve the control problem of fluid parameter perturbation and external interference conditions in the motion of underwater vehicles, corresponding nonlinear control methods are needed.

[0052] When the uncertain part of the control system is an unknown function, the neural network control method can be used. Since neural network control does not rely on the precise information of the controlled object and does not require the establishment of fuzzy rules, it can be applied to the field of underwater vehicle attitude control. The research on the combination of neural network and adaptive control has become a new branch of intelligent control. Adaptation has strong robustness, and neural network has good self-learning function and good fault tolerance. Neural network adaptive control has a strong advantage because it better integrates the advantages of both. Therefore, applying adaptive neural network to underwater vehicle control can improve the performance of underwater vehicle control system.

[0053] The adaptive neural network used in the early underwater vehicle control could not guarantee the stability of the system because the adjustment of the neural network weights was derived from the defined minimization objective function. With the widespread application of Lyapunov stability theory, adaptive neural networks based on Lyapunov stability theory are widely used in underwater vehicle attitude control systems.

[0054] The adaptive laws of weight adjustment of neural network control schemes are derived based on Lyapunov stability theory, which avoids a lot of complex calculations, and can ensure the global stability of the system. It has good robustness to the uncertainty of underwater vehicle models, parameter uncertainty and external disturbances. Therefore, it is one of the more feasible adaptive neural network control schemes for underwater vehicle control. However, since most adaptive neural networks use multi-layer neural network structures, these neural network structures are complex and have low computational efficiency. In addition, they generally use fixed-structure neural networks, which have the phenomenon of lack of neurons and redundancy. Gaussian RBF neural networks and the recently studied convex cosine RBF neural networks and variable structure neural network technologies provide a broader space for adaptive neural networks in the direction of underwater vehicle control.

[0055] Since the neural network is a universal approximator, that is, as long as there are enough parameters, it can approximate any smooth function with arbitrary accuracy. Then, we can use neural networks to model unknown functions and estimate network hyperparameters to achieve the purpose of adaptive control. This is also the route currently adopted by neural network-based algorithms such as deep learning and reinforcement learning. However, current control theory attaches great importance to stability analysis, and for deep neural networks, the stability of closed-loop systems is usually difficult to prove because deep networks cannot ensure convergence to the optimal point. Then, neural network adaptive control requires proof of stability. However, most of the proofs are currently limited to single-layer networks, that is, there is only one layer of hyperparameters. In the field of control, we usually use radial basis functions (RBF) to complete the approximation of arbitrary functions. The radial basis function usually uses a Gaussian function as the kernel function (activation function), which also has universal approximation properties. Most importantly, it is single-layer, so it can be guaranteed by stability theory.

[0056] The neural network adaptive method of single-channel equations can solve the problems of simple systems. However, many of the models we usually need to deal with are of second order or above, and the adaptive control design in this case is more universal.

[0057] The adaptive control method is a design method that estimates the uncertain controlled object parameters (or equivalently, the corresponding controller parameters) online based on the measured signals in response to the changes in the controlled object parameters, and uses the estimated values ​​of the parameters in the control calculation.

[0058] In the 1950s, the Massachusetts Institute of Technology (MIT) proposed the first adaptive controller. Subsequently, various adaptive control methods emerged one after another. Among them, the current mainstream adaptive control methods mainly include model reference adaptive control and self-correcting control.

[0059] Early adaptive control was based on the MIT gradient method, which used a gradient descent method similar to that in signal processing. Currently, adaptive control in control theory focuses on stability, and Lyapunov theory plays an important role in it.

[0060] Usually, in actual control system design, adaptive control methods are often combined with many control methods, such as adaptive PID control, adaptive sliding mode control method and neural network adaptive control method, etc.

[0061] The concept of adaptive control is quite broad and can be divided into feedforward adaptive control, feedback adaptive control, model reference adaptive control and self-correcting control.

[0062] Model Reference Adaptive Control (MRAC) Figure 1As shown in Figure 1, in this control framework, a so-called "reference model" can be artificially constructed to characterize the nominal closed-loop system control performance. Model reference adaptive control is intended to obtain a dynamically adjusted feedback control law so that the closed-loop control performance of the system can be consistent with that of the reference model.

[0063] Depend on Figure 1 It can be seen that the model reference adaptive control system contains two different loops, in which the inner loop consists of the controlled object and the controller, and is connected in parallel with the reference model; while the outer loop is an adaptive mechanism that adjusts the controller parameters. Since the reference signal applied to the adjustable controller also acts on the reference model, the control performance index expected by the system can be determined by the output or state of the reference model.

[0064] Transient response refers to the process of the system output changing from the initial state to the stable state under the action of a typical signal input. Transient response is also called dynamic response, transition process or transient response. It is usually determined by the dynamic characteristics of the system, including factors such as the inertia and lag of the system.

[0065] In underwater vehicle control systems, transient response refers to the process in which the vehicle's attitude changes from one stable state to another, such as the heading changes from 0° to 90°; the depth changes from 0 meters to 5 meters, etc. During these attitude changes, transient behaviors such as overshoot, attenuation and oscillation will occur.

[0066] A control system with good transient response should react immediately to changes in signals or attitudes. Once the input signal stops, the response should also stop immediately without problems such as oscillation and lag. In addition, transient response is different from steady-state response. Steady-state response is the response of the system to the input signal after reaching a steady state, usually manifested as a stable component, such as a step function and a periodic function.

[0067] Based on the above control system analysis, this application proposes an underwater vehicle model reference adaptive control system and method based on transient performance optimization. The system can compensate for the uncertainty of the actuator, suppress system oscillation, reduce experimental costs, realize rapid and stable error-free control of the underwater vehicle, alleviate the contradiction between the rapidity and error of the control system, achieve a stable error-free control effect, and complete the transient performance optimization of the control system of the underwater vehicle.

[0068] Underwater vehicle model reference adaptive control is suitable for the control system of autonomous underwater vehicles with time-varying, nonlinear and parameter uncertain characteristics.

[0069] In the traditional PID control of underwater vehicles, we need to adjust the depth and heading of the vehicle. The usual method is to directly give a set value. For example, we require the depth of the vehicle to change from 0 meters to 5 meters, such as Figure 2As shown. This is actually equivalent to a step response of a depth channel. Since the input command is discontinuous, the system response usually has overshoot and oscillation. The depth channel of the underwater vehicle usually uses an integral link to eliminate the steady-state error, which will bring greater oscillation and lag. This is the transient response of the system we mentioned above. The reason is due to the influence of the step signal. We can reduce this influence through the reference model. In other words, we used to track a discontinuous step signal, but now we can track the continuous signal generated by the reference model, which asymptotically approaches the step signal, thereby completing the transient performance optimization.

[0070] Normally, the design of high-performance underwater vehicles should go through processes such as fluid simulation optimization and water tunnel fluid parameter testing to determine the main fluid parameters, establish a high-precision mathematical model of the vehicle, and then design a controller to carry out simulation work. However, the cost of water tank testing remains high. In order to save money, people hope to design a controller with good performance even when the main fluid parameters are not very accurate. This can be achieved by using the model reference adaptive control described in the present invention.

[0071] When underwater vehicles are in navigation control, they are affected by interference, including changes in rudder zero bias and rudder force coefficient. Low-cost vehicles do not undergo hydrodynamic tests due to financial considerations, and parameter errors are large, affecting the control effect. The use of adaptive methods can offset the impact of parameter uncertainty, but the adaptive process is a process of changing the parameters to be estimated, which affects the transient response of the system and causes oscillations in the transient response. In many cases, this is undesirable.

[0072] The system and method proposed in this application can effectively suppress the transient response oscillation phenomenon in the model reference adaptive control process.

[0073] The technical solution of underwater vehicle model reference adaptive control system and method based on transient performance optimization includes:

[0074] 1. High-level system model construction

[0075] Consider the following system:

[0076]

[0077] Among them, u is the system input signal, a and b are system parameters, and x is the system state variable. is the derivative of the system state variable.

[0078] Usually, the aircraft model that needs to be processed is of second order or higher, and the adaptive control design in this case is more universal. From equation 1, we consider the system model to be a high-order system equation with uncertainty:

[0079]

[0080] Among them, A and B are known and controllable matrices, d(x,t) is an uncertain term, Λ>0 is a diagonal matrix, and satisfies the Lipschitz condition:

[0081]

[0082] Where l is a constant used to describe the uncertainty of controllability; ||·|| is the norm calculation.

[0083] 2. Design ideas

[0084] It can be seen that the high-order system (2) is more complex, with both state uncertainty and controllability uncertainty. It is not difficult to see that the state uncertainty acts on the u channel. We call this uncertainty matching uncertainty. We can completely suppress the matching uncertainty by designing the control input u, which is the design idea of ​​the present invention.

[0085] 3. Model Reference Adaptive Controller Design

[0086] (1) Design the closed-loop control variable u according to the nominal model n

[0087] Since there are many uncertainties in the system, the model reference model is introduced to simplify the problem. In the original system, let Λ=I, d(x, t)=0 to get the nominal system. Equation (2) can be written as:

[0088]

[0089] If y is the system output, y r is the output set value, then the design goal is to adjust the system output to the set value y r ,Right now:

[0090]

[0091] Among them, C is a known matrix.

[0092] Design closed-loop control according to the nominal system, and the nominal system control law is:

[0093] u n =-K 1 x+K 2 y r (6)

[0094] Among them, K 1 To adjust the closed-loop poles, make:

[0095]

[0096] is the Hurwitz matrix (characteristic root is in the right half plane), K 2 The selection principle is based on y r Adjust to ensure the closed-loop system steady-state output y(∞) = y r From the closed-loop equation of the system, substitute equation 6 and equation 7 into equation 4, and from the closed-loop equation of the system, we get:

[0097]

[0098] When t→∞, x→x final , then:

[0099] 0=A r x final +BK 2 y r (9)

[0100] x final =-A -1 BK 2 y r (10)

[0101] x final ,y final is the final value of the system. From formula 5, we get:

[0102] y final =Cx final =-CA -1 BK 2 y r (11)

[0103] We hope that final =y r , then K 2 The selection must meet the following requirements:

[0104] -CA -1 BK 2 =-I (12)

[0105] So far, according to the designed K 1 ,K 2 , assuming B r =BK 2 , substituting into equation 8, the nominal system or reference model can be expressed as:

[0106]

[0107] In order to distinguish it from the original system, we represent the state variable of the reference model as x r .

[0108] (2) Neural network design additional control variable u a

[0109] like Figure 3 As shown, according to the consistency approximation principle of neural network, we can select the basis function Θ 0 (x) and the weight matrix W 0 , so that the uncertainty can be expressed as

[0110]

[0111] Where T is the matrix transpose.

[0112] The basis function can be a set of Gaussian functions or can be selected according to the specific model structure. Then the closed-loop control quantity of the system can be expressed as the sum of the nominal control and an additional control, that is,

[0113] u=u n +u a (15)

[0114] Substitute equations (14) and (15) into system equation (2) and perform the following processing:

[0115]

[0116] Using Ax+Bu n =A r x+B r y r , and rewrite the third term on the right side of the equal sign, and set W = [W 0 ,I-Λ -1 ] T ,Θ=[Θ 0 ,u n ] T , I is the unit matrix, forming a new extended basis function:

[0117]

[0118] Therefore, the closed-loop state equation of the system is finally expressed as:

[0119]

[0120] Combine the above formula with the reference model Subtract, and set have to:

[0121]

[0122] It can be seen that if the estimated weight is The additional control quantity should be defined as:

[0123]

[0124] Assume the weight estimation error is Then we have:

[0125]

[0126] (3) Neural Network Weight Update Law

[0127] Construct the Lyapunov function:

[0128]

[0129]

[0130] Because A r is a Hurwitz matrix, so for any Q>0, there exists P>0 such that:

[0131]

[0132] We take it again.

[0133]

[0134] Then Equation 23 can be simplified as:

[0135]

[0136] According to Barbalat's theorem Therefore, the system is asymptotically stable, and Bounded, only when the system meets the continuous incentive conditions,

[0137] Introducing a reference model allows the system to track a smooth signal instead of a step signal, which can improve the transient response of the system and reduce overshoot and oscillation. However, we note that due to the adaptive weight The introduction of the adaptive update of parameters and the adjustment of the control quantity will produce a coupling effect, which will also have an adverse effect on the transient performance of the system.

[0138] 4. Transient performance enhanced reference adaptive control quantity v

[0139] From formula 21 we can see that This term reduces the transient performance of the closed-loop system. In order to reduce the transient performance degradation caused by adaptation, we can compensate for it by adding another term v. That is, u = u n +u a +v, then equation 16 can be written as:

[0140]

[0141] Similarly, we select

[0142]

[0143] Then (27) is simplified to:

[0144]

[0145] According to formula 29, v is used to suppress At this point we make the following changes to the reference model:

[0146]

[0147] make but:

[0148]

[0149] So far, we have concluded that if we order:

[0150]

[0151] Bringing in (29) can offset The closed-loop system equation is:

[0152]

[0153] But the reality is that we cannot directly obtain We can use the usual signal processing method to introduce filtering get Substituting 32 into the estimate, we get:

[0154]

[0155] Where μ is the filter time constant. Obviously, as μ increases, Equation 34 is closer to the ideal output. s is the complex variable in the Laplace transform.

[0156] So far, if Figure 4 As shown, the system control quantity u=u n +u a +v, nominal control quantity u n , additional control quantity u a and the transient performance enhancement control quantity v have been designed.

[0157] Below we give a specific implementation use case of adaptive heading control for underwater vehicles:

[0158] Consider the aircraft heading channel:

[0159]

[0160] Where ψ and ω are the heading angle and heading angular velocity, respectively, b is the estimated value of the control coefficient, which is usually different from the true coefficient by a factor; we use the unknown number λ to represent it. β is the sideslip angle, f(ω,β,v t ) are uncertain terms related to angular velocity, sideslip angle and speed. Since we do not have accurate water tank test data, we use adaptive technology to compensate for these terms.

[0161] In addition, in the system model, the servo has an unknown mechanical zero point. Although we can compensate the mechanical zero point of the servo more accurately, this work is usually time-consuming and laborious. For the sake of simplicity, the mechanical zero point Δ is also included in the model. Since the term related to Δ is a constant, we combine it with f(·) to form the disturbance term d(ω,β,v t ), then the original state equation is transformed into

[0162]

[0163] Equation 36 is in the form of a system state equation, and the model reference adaptive control of the present invention is used to complete this control problem containing multiple uncertainties.

[0164] Build the system model:

[0165] Define the system state variables as

[0166] x=[ψω] T (37)

[0167] Then the state equation of the high-order system can be expressed as:

[0168]

[0169] Among them, select B=[0,5] T is the nominal value of the control matrix, η = 0.667 means that the actual control action is only 2 / 3 of the nominal value, which is unknown to the controller designer, and u is the rudder angle. Due to the experimental conditions, the additional mass of the spacecraft cannot be known, and because the rudder has an unknown mechanical zero point, we represent all these uncertainties with d(x). Since the differential of the heading angle is the heading angular velocity, the above uncertainties only appear in the second differential equation, that is, in the channel acting on u. We call such uncertainties matching uncertainties.

[0170] d(x)=W T Θ(x) (39)

[0171] Considering that d(x) is related to the rudder angle zero bias (constant) and angular velocity, the basis function is selected as:

[0172] Θ(x)=[1,x2 ] T (40)

[0173] The nominal system model is:

[0174]

[0175] The nominal controller can be designed using pole placement or LQR method, taking:

[0176] Q=I 2 ,R=5

[0177] K can be obtained by using the following Matlab function LQR and (12) 1 ,K 2 :

[0178]

[0179] The nominal control u n for:

[0180] u n =-[0.44722,0.6155]x+0.4472y r (43)

[0181] Substitute into the closed-loop state equation in (13) B r =BK 2 In the result:

[0182]

[0183] Easy to know A r is the Hurwitz matrix, and the system is stable.

[0184] uControl law design:

[0185] Since d(x) includes the rudder angle zero bias and the additional mass of the heading channel, in the simulation W 0 =[-2*π / 180,-1.5], in Α r T P+PA r = -Q, since Q = I 2 , so we can use the matlab command P = lyap (Ar', eye (2)) to solve

[0186]

[0187] In (Equation 25) Select γ=15 and set P,B,e=xx r Substitute into equation 25 and note that in equation 40, Θ(x) = [1, x2 ] T , so we have completed the design of the adaptive control rate. In formula 34, take μ = 10.

[0188] Simulation results analysis:

[0189] Since the rudder angle has a 2° zero-position deviation, η=0.667, it means that the actual rudder effect is only 2 / 3 of the estimated value. At the same time, the additional mass of the heading channel is unknown, that is, we do not know the inertial lag of the channel. In this case, it is difficult to design an ordinary controller, and PID on-site debugging also requires very rich on-site experience to complete.

[0190] By adopting the reference model adaptive control system of the present invention, we can compensate for the servo zero point and other uncertainties online, and finally achieve better control effect.

[0191] Simulation Results Figure 5 and Figure 6 , the heading instruction, the reference model output heading angle (navigation attitude) and the actual system output heading angle are given.

[0192] It can be seen that under the same parameters, both adaptive controllers can roughly complete the task of heading tracking. However, due to the influence of the rudder angle zero bias, the transient tracking performance of the ordinary model reference adaptive control is reduced. Figure 5 The angle and angular velocity curves are shown in the figure.

[0193] After adopting the transient performance optimization method of the present invention, by introducing the transient performance enhancement compensation control quantity v, that is, u = u n +u a +v, offsetting the effect of parameter estimation error on transient performance and better tracking the reference model. Figure 6 shown.

[0194] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application is described in detail with reference to the embodiments, a person skilled in the art should understand that any modification or equivalent replacement of the technical solution of the present application does not depart from the spirit and scope of the technical solution of the present application and should be included in the scope of the claims of the present application.

Claims

1. A model reference adaptive control system for underwater vehicles based on transient performance optimization, characterized in that: The system comprises: The system model of the system is: where x = [ψω] T is the system state variable, ψ is the heading angle of the underwater vehicle, ω is the heading angular velocity of the underwater vehicle; T is the vector transpose; is the derivative of the system state variable, u(t) is the system input signal, that is, the rudder angle of the actuator; A and B are controllable matrices; d(x,t) is the uncertain term; t is time; Λ>0 is a diagonal matrix that satisfies the Lipschitz condition: Where l is a constant; ||·|| is the norm calculation; The reference model of the system is: Among them, x r is the system state variable of the reference model, is the derivative of the system state variable of the reference model, y r Set the value for the output; is the Hurwitz matrix with characteristic roots in the right half plane; K1 is the regulation closed-loop pole; B r =BK2; the selection principle of parameter K2 is based on y r Adjust to ensure that the closed-loop system steady-state output is y r ; The system control quantity u of the system is expressed as: u=u n +u a +v Among them, u n is the nominal control quantity; u a is the additional control quantity; v is the transient performance enhancement control quantity.

2. The underwater vehicle model reference adaptive control system based on transient performance optimization according to claim 1, characterized in that: The nominal control quantity u n for: you n =-K1x+K2y r 。 3. The underwater vehicle model reference adaptive control system based on transient performance optimization according to claim 1, characterized in that: The additional control amount u a for: Among them, Θ(x)=[Θ0(x),u n ] T , Θ0(x) is the basis function; is the estimated weight.

4. The underwater vehicle model reference adaptive control system based on transient performance optimization according to claim 3 is characterized in that: The basis function Θ0(x) is a set of Gaussian functions.

5. The underwater vehicle model reference adaptive control system based on transient performance optimization according to claim 1, characterized in that: The transient performance enhancement control amount v is: in, μ is the filter time constant; s is the complex variable in Laplace transform.

6. The underwater vehicle model reference adaptive control system based on transient performance optimization according to claim 1, characterized in that: The parameter K2 is selected to satisfy: -CA -1 BK2=-I Where I is the identity matrix; C is a matrix that satisfies: Among them, y is the system output value.

7. A method for underwater vehicle model reference adaptive control based on transient performance optimization, implemented based on the system of any one of claims 1 to 6, the method comprising: Build system models and reference models; Determine the basis function according to the factors affecting the underwater vehicle track, calculate the nominal control quantity, additional control quantity and transient performance enhancement control quantity, and determine the adaptive control system; By inputting the rudder angle of the actuator into the system model, the navigation attitude of the underwater vehicle can be output, completing the model reference adaptive control of the underwater vehicle.

Images