CN102819220A - Adaptive control method of autopilot of ship - Google Patents

Abstract

The invention discloses an adaptive control method of an autopilot of a ship and relates to the technical field of automatic control of the ship. The method comprises the steps of: firstly, carrying out wave model identification; updating the upper limit and the lower limit of a controller pipeline so as to calculate controlled quantity; further judging that if the wave model is in a still water area, the model validation is not passed; then, carrying out ship model identification; updating a controller; and carrying out the next control. According to the invention, the self-setting function of the autopilot is automatically finished without personnel to participate and can automatically identify the sea state grade, so that adaptive heading control under wave interference is achieved, and the control performance of the heading of the ship is improved.

Description

The Marine Autopilot self-adaptation control method

Technical field

The present invention relates to boats and ships automatic control technology field, be specifically related to a kind of Marine Autopilot self-adaptation control method.

Background technology

Nineteen twenty-two, the course autopilot obtained using first in ship course control, and the Marine Autopilot synoptic diagram is as shown in Figure 1.Because the maturation of PID control technology in industry used, in the long duration, the PID autopilot has occupied leading position in the ship course control field.Commercial course autopilot product mainly is distributed in the flourishing area of shipping industries such as Japan, Europe, Canada; Like the G-PILOT 3380 of Britain NAVMANW company, the AP50 of SIMRAD company; The Admiral-P3 of Canadian ComNav company, series of products of Japanese Marol company etc.In recent years, domestic seafaring is flourish, and successively many companies have released the autopilot product, and like sea, Beijing blue letter HLD-SC100, Jinzhou profit reaches THD2001A, the HD-6AT1 of middle ship heavy industry 707 research institutes (Jiujiang).

Yet one of problem that exists at present is that adjusting of PID coefficient generally adopts method of trial and error to adjust by manual observation control effect.Because model linearizing working point in course is index with the speed of a ship or plane, thus need be under the different speed of a ship or plane tuning PID coefficient, workload is bigger.And when dispatching from the factory, most of autopilot parameter solidifies, when environment changes such as boat-carrying, boat territory, and the control effect in the time of can not guaranteeing to design.Therefore, design a kind of course autopilot and can reduce crewman's labour intensity greatly, and improve the control effect with self-setting function.

In addition, boats and ships receive wave and disturb greatlyyer when surface navigation, and the steering wheel frequent movement but can't bring positive control effect, and produces bigger mechanical noise, and the crewman also will stand ear-piercing noise when enduring the hardship that wave jolts to the fullest extent.Method at present commonly used is, beats the rudder number of times and is manually switched to the pattern of going as course for reducing, and takes the method for increase control dead area on the basis of PID autopilot usually.This method tends to cause driftage, and is difficult to reduce effectively beat the rudder number of times.Therefore, design a kind of according to sea situation can adaptive setting the course keep adaptive rudder to be necessary.

Domestic and international ship course control technology is summed up as follows:

A, domestic boats and ships adaptive rudder technology are almost blank with product;

B,, wave needs the manual switching direction controller under disturbing;

The course of C, use control dead area method keeps autopilot control effect relatively poor.

Summary of the invention

The object of the present invention is to provide a kind of self-adaptive ship autopilot control method; Realization need not the autopilot self-setting function that personnel participate in, adjust and accomplish automatically; And, realize the course adaptive control under the wave interference with automatic identification sea situation grade, improve the control performance of ship course.

The Marine Autopilot self-adaptation control method is specially:

(1) measure boats and ships in the course of current time k ψ (k);

(2) make up ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), z ^-1Be unit factor retardation time, A (z ^-1) and B (z ^-1) be known multinomial coefficient, δ (k-1) is the controlled quentity controlled variable of moment k-1 ,-20＜e (k)＜20 finds the solution ship model and obtains multinomial coefficient C (z ^-1);

(3) with multinomial coefficient C (z ^-1) substitution ship model A (z ^-1) ψ (n)=B (z ^-1) δ (n-1)+C (z ^-1) e (n), n=k-n ₀..., k, 5 ₀Uncorrelated random series e (n) is found the solution in≤n0≤200, calculates covariance

In conjunction with the classification of seas of the covariance of setting up in advance, confirm the course pipeline upper limit according to the classification of seas of current time k with the mapping relations demarcation current time k of classification of seas

And lower limit

Output controlled quentity controlled variable δ (k), wherein $C=\left[\begin{array}{c}I\\ -I\\ C_{I/\mathrm{\Δ}}\\ -C_{I/\mathrm{\Δ}}\\ H\\ -H\end{array}\right],d_k=\left[\begin{array}{c}\mathrm{\Δ}\stackrel{\‾}{\mathrm{\δ}}\\ -\mathrm{\Δ}\underset{\‾}{\mathrm{\δ}}\\ \stackrel{\‾}{\mathrm{\δ}}-L{\mathrm{\δ}}_{k-1}\\ -\underset{\‾}{\mathrm{\δ}}+L{\mathrm{\δ}}_{k-1}\\ \stackrel{\‾}{\mathrm{\ψ}}-\mathrm{P\Δ}\underset{\←}{\mathrm{\δ}}-Q\underset{\←}{\mathrm{\ψ}}\\ -\underset{\‾}{\mathrm{\ψ}}+\mathrm{P\Δ}\underset{\←}{\mathrm{\δ}}+Q\underset{\←}{\mathrm{\ψ}}\end{array}\right],$

H＝C _D ^-1C _B，P＝C _D ^-1H _B，Q＝-C _D ^-1H _D，

D(z ^-1)＝ΔA(z ^-1)，

C _B, H _BBe multinomial coefficient B (z ^-1) Toeplitz and Hankle matrix,

C _D, H _DBe multinomial coefficient D (z ^-1) Toeplitz and Hankle matrix,

C _{The I/ Δ}Be the Toeplitz matrix of polynomial expression I/ Δ,

is respectively steering wheel increment rudder angle upper physical limit and lower limit

is respectively the upper physical limit and the lower limit of steering wheel rudder angle amplitude

Δ＝1-z ^-1，

$f=-H^T{W}_{\mathrm{\ψ}}({\underset{\→}{\mathrm{\ψ}}}^{*}-\mathrm{P\Δ}\underset{\←}{\mathrm{\δ}}-Q\underset{\←}{\mathrm{\ψ}}),$ S＝H ^TW _ψH+W _δ， $L=\left[\begin{array}{c}1\\ 1\\ \·\\ \·\\ \·\\ 1\end{array}\right],$

plays the rudder angle value collection of one period continuous historical juncture of counting forward for a last moment k-1;

plays the course value collection of the continuous historical juncture of section of counting forward for current time k;

equates with element number in

The rudder angle value collection in one period moment of future that

counts for the current time k that finds the solution backward

The course prediction value collection in one period moment of future that

counts for current time k backward;

is the following course set-point collection one to one constantly in

Diagonal angle, course weighting matrix W _ψBe unit matrix,

W _δBe diagonal angle, course weighting matrix, the value on its diagonal line gets 0.1～0.5,

I is the unit diagonal matrix, and subscript T is a transposition;

(5) if current classification of seas be zero and covariance

greater than the predictive error threshold value; Then get into step (6); Otherwise, get into next and control constantly;

(6) with multinomial coefficient C (z ^-1) substitution ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), find the solution multinomial coefficient A (z ^-1) and B (z ^-1), and then update controller, get into next and control constantly.

Said step (2) is found the solution ship model according to following mode and is obtained multinomial coefficient C (z ^-1):

I. calculate ξ (k)=A (z ^-1) ψ (k)-B (z ^-1) δ (k-1)-(C ₁z ^-1+ C ₂z ^-2+ C ₃z ^-3+ C ₄z ^-4) e (k) renewal

Structure

Ii. ∑ being carried out singular value decomposes ${\stackrel{\‾}{U}}_k\left[\begin{array}{c}{\stackrel{\‾}{D}}_k\\ 0\end{array}\right]{\stackrel{\‾}{V}}_k^T=\mathrm{Svd}\left(\mathrm{\Σ}\right)$ Confirm matrix

Svd () is the svd function;

Iii. upgrade $U_k=U_{k-1}{\stackrel{\‾}{V}}_k^T,$ $D_k={{\stackrel{\‾}{D}}_k}^{-1};$

Iv. definition

Forms of characterization do ${\widehat{\mathrm{\&theta;}}}_k={[{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HDA00001926437800013.png"\; state="\{\{state\}\}">c</figure-callout>}}_1,c_2,c_3,c_4]}^T,$ Upgrade

V. foundation

Evaluator coefficient C (z ^-1)=1+C ₁z ^-1+ C ₂z ^-2+ C ₃z ^-3+ C ₄z ^-4

Said step (6) is found the solution ship model according to following mode and is obtained multinomial coefficient A (z ^-1) and B (z ^-1):

1) upgrades

and construct

2) ∑ being carried out singular value decomposes ${\stackrel{\&OverBar;}{U}}_k\left[\begin{array}{c}{\stackrel{\&OverBar;}{D}}_k\\ 0\end{array}\right]{\stackrel{\&OverBar;}{V}}_k^T=\mathrm{Svd}\left(\mathrm{\&Sigma;}\right)$ Confirm matrix

Svd () is the svd function;

3) upgrade $U_k=U_{k-1}{\stackrel{\&OverBar;}{V}}_k^T,$ $D_k={{\stackrel{\&OverBar;}{D}}_k}^{-1};$

4) definition

Forms of characterization do ${\widehat{\mathrm{\&theta;}}}_k={[a_1,a_2,b_0,b_1]}^T,$ Upgrade

Foundation

Evaluator coefficient A (z ^-1)=1+a ₁z ^-1+ a ₂z ^-2And B (z ^-1)=(b ₀+ b ₁z ^-1) z ^-1With existing autopilot technology, the present invention has following characteristics:

A, do not need manual intervention, accomplish the identification of boats and ships parameter and adjusting certainly of autopilot automatically;

B, can on-line identification wave model, demarcate sea situation progression, do not need manual switchover to course retainer;

C, through identification sea situation progression, set the pipeline upper bound and lower bound, realize that boats and ships disturb course down to control at wave, have overcome the driftage that cause in the dead band, and have reduced effectively and beaten the rudder number of times.

Description of drawings

Fig. 1 is the Marine Autopilot structural drawing;

Fig. 2 is non-linear " pipeline " control synoptic diagram;

Fig. 3 is an identification experimental design synoptic diagram;

Fig. 4 is a core control method process flow diagram of the present invention.

Embodiment

The Marine Autopilot structured flowchart is as shown in Figure 1, and the present invention is the technological core of ship course controller, and the instruction of accomplishing autopilot produces.

(1) design course Model Distinguish test

That become when the controller of feedback channel is or have non-linearly, closed-loop system is cognizable.Therefore the present invention adopt as the identification experimental design method.Update controller immediately behind the identification model, this moment, system was cognizable.

(2) course model parameter and wave identification of Model Parameters are switched

At first when boat-carrying changes greatly, start the boats and ships discriminating function in the waters of relatively placidity, the identification ship model, identification is closed discriminating function after accomplishing.In underway, adaptive rudder can indicate classification of seas and identification state for crewman's reference automatically.

(3) course Model Distinguish and control

Autopilot is powering on constantly, correlated variables is carried out customary initialization (use currency to fill when comprising that the historical juncture does not have value; Load boats and ships initial model value etc.); Carry out the wave Model Distinguish then, and the update controller pipeline upper limit

and lower limit

calculation control amount.Further judge, then carry out the ship model identification if hydrostatic territory and modelling verification are not passed through, and update controller (referring to Fig. 3), get into next and control constantly.

Because the present invention has higher robotization, more than the enforcement of each step do not need manual intervention, realized motor-driven maintenance and the control with retentive control of ship course fully.This adaptive rudder is installed, need upgrade the original autopilot equipment of boats and ships, only need on the basis of existing procucts, to carry out software upgrading.Therefore, installation cost is low, and benefit is big.

On-line identification ship course model of the present invention; And model output and actual measurement course compared, if error is bigger, the expression identification model can't satisfy actual requirement; Identification algorithm carries out identification and model parameter is disposed autopilot, reaches the adaptation function of autopilot course control.On the basis of course Model Distinguish, demarcate classification of seas through identification wave model, and with it foundation that is provided with as " pipeline " control bound, thereby realize that wave disturbs the adaptive control in course down.

(1) ship model identification ultimate principle of the present invention

Known boats and ships ARMAX model,

A(z ^-1)ψ(k)＝B(z ^-1)δ(k-1)+C(z ^-1)e(k) (1)

In the formula (1),

δ (k-1) is last one ship model yaw rudder input constantly;

ψ (k) measures the course for current time;

E (k) is uncorrelated random series;

A (z ^-1), B (z ^-1) be ship model parameter time series

A(z ^-1)＝1+a ₁z ^-1+a ₂z ^-2

B(z ^-1)＝(b ₀+b ₁z ^-1)z ^-1

Wave model parameter time series

C(z ^-1)＝1+C ₁z ^-1+C ₂z ^-2+C ₃z ^-3+C ₄z ^-4

Can adopt least square method, Li Baige-Ma quart (LM) and output-error method (OE) or the like method to find the solution above-mentioned model.Ship course Model Distinguish technology of the present invention is incorporated into Singular Value Decomposition Using in the least square identification, accomplishes the identification of ship model parameter, and disposes generalized predictive controller automatically.Singular Value Decomposition Using has good numerical characteristic; Eliminated the existence of above numerical evaluation problem; Maximum speed of convergence and the identification precision that has reduced theoretical least square identification algorithm in practical application, the recursive algorithm of ARMAX model identification ship model parameter effectively under the different speed of a ship or plane.The model parameter of identification boats and ships in the hydrostatic territory, and storage.Under wave disturbs, the on-line identification classification of seas, the crewman is given in indication.Specify below:

1) specify forgetting factor μ (0.95＜μ＜1), initial value generally is taken as U ₀=I, D ₀=10 ⁴I,

K ₀=0, other initial value is initialized as 0;

2) upgrade

and construct

3) ∑ being carried out singular value decomposes ${\stackrel{\&OverBar;}{U}}_k\left[\begin{array}{c}{\stackrel{\&OverBar;}{D}}_k\\ 0\end{array}\right]{\stackrel{\&OverBar;}{V}}_k^T=\mathrm{Svd}\left(\mathrm{\&Sigma;}\right);$

4) upgrade $U_k=U_{k-1}{\stackrel{\&OverBar;}{V}}_k^T;$

5) upgrade $D_k={{\stackrel{\&OverBar;}{D}}_k}^{-1};$

6) upgrade

7) upgrade and forward 2 to).

Wherein, at identification ship model A (z ^-1), B (z ^-1) time

θ＝[a ₁，a ₂，b ₀，b ₁] ^T

y _k＝ψ(k)

At wave MODEL C (z ^-1) time

θ＝[c ₁，c ₂，c ₃，c ₄] ^T

y _k＝ξ(k)

ξ(k)＝A(z ^-1)ψ(k)-B(z ^-1)δ(k-1)-(C ₁z ^-1+C ₂z ^-2+C ₃z ^-3+C ₄z ^-4)e(k)

Through above renewal, can identification obtain ship model in the hydrostatic territory.At the water surface that wave disturbs, can be through identification C (z ^-1) polynomial expression demarcates the wave superfine.

(2) ultimate principle of course of the present invention " pipeline " control

" pipeline " is controlled to be one of the control model of the Prediction and Control Technology of belt restraining.The present invention's " pipeline " control is applied to the Marine Autopilot control technology, and the course is constrained in given " pipeline "." pipeline " is controlled at each feasible region and is equivalent to the submodel control system, and when constraint condition changed, " pipeline " was controlled at each feasible region switching and is equivalent to multi-model control.When the course near set-point, the feasible region system-gain is less, controller is stronger to the wave Disturbance Rejection; When the controller switching caused restrained boundary, the feasible region system-gain was bigger, and the direction controller tracking power strengthens." pipeline " is controlled at and carried out effective balance design between noise inhibiting ability and the tracking power, and this balance is carried out under this control model of " pipeline " control, so the course control under wave disturbs can obtain to control preferably effect.The PREDICTIVE CONTROL of belt restraining " pipeline " course control technology background is among the present invention: under the bigger environment of noise; Output to controlling object there is no need maybe can't realize accurate control (tight control); But it is constrained in the middle of a certain interval, interval bound is a constant, is the both sides parallel lines on time shaft; Be called visually " pipeline " control, as shown in.Boats and ships disturb course control down at wave, cause steering wheel significantly, high-frequency invalid action, the present invention will " pipeline " control introduces the course control of boats and ships, in the process that goes as course the minimizing steering wheel is moved to have great importance.Not only saved fuel but also reduced mechanical noise.

When there was constraint in generalized predictive control, there were not analytic solution in the generalized predictive control rule, is in the nature the optimization problem of finding the solution belt restraining, and its canonical form is:

$\left\{\begin{array}{c}\underset{\mathrm{\&Delta;}\underset{\&RightArrow;}{u}}{\min }J=\mathrm{\&Delta;}\underset{\&RightArrow;}{\mathrm{\&delta;}}\mathrm{S\&Delta;}{\underset{\&RightArrow;}{\mathrm{\&delta;}}}^T+2f^T\mathrm{\&Delta;}\underset{\&RightArrow;}{\mathrm{\&delta;}}\\ \mathrm{C\&Delta;}\underset{\&RightArrow;}{\mathrm{\&delta;}}-d_k\&le;0\end{array}\right.$

Wherein

$C=\left[\begin{array}{c}I\\ -I\\ C_{I/\mathrm{\&Delta;}}\\ -C_{I/\mathrm{\&Delta;}}\\ H\\ -H\end{array}\right],d_k=\left[\begin{array}{c}\mathrm{\&Delta;}\stackrel{\&OverBar;}{\mathrm{\&delta;}}\\ -\mathrm{\&Delta;}\underset{\&OverBar;}{\mathrm{\&delta;}}\\ \stackrel{\&OverBar;}{\mathrm{\&delta;}}-L{\mathrm{\&delta;}}_{k-1}\\ -\underset{\&OverBar;}{\mathrm{\&delta;}}+L{\mathrm{\&delta;}}_{k-1}\\ \stackrel{\&OverBar;}{\mathrm{\&psi;}}-\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}-Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}\\ -\underset{\&OverBar;}{\mathrm{\&psi;}}+\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}+Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}\end{array}\right],$

H＝C _D ^-1C _B，P＝C _D ^-1H _B，Q＝-C _D ^-1H _D，

D(z ^-1)＝ΔA(z ^-1)，

C _B, H _BBe multinomial coefficient B (z ^-1) Toeplitz and Hankle matrix,

C _D, H _DBe multinomial coefficient D (z ^-1) Toeplitz and Hankle matrix,

C _{The I/ Δ}Be the Toeplitz matrix of polynomial expression I/ Δ,

is respectively steering wheel increment rudder angle upper physical limit and lower limit

is respectively the upper physical limit and the lower limit of steering wheel rudder angle amplitude

Δ＝1-z ^-1，

$f=-H^T{W}_{\mathrm{\&psi;}}({\underset{\&RightArrow;}{\mathrm{\&psi;}}}^{*}-\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}-Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}),$ S＝H ^TW _ψH+W _δ， $L=\left[\begin{array}{c}1\\ 1\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ 1\end{array}\right],$

plays the rudder angle value collection of one period continuous historical juncture of counting forward for a last moment k-1;

plays the course value collection of one period continuous historical juncture of counting forward for current time k;

equates with element number in

The rudder angle value collection in one period moment of future that

counts for the current time k that finds the solution backward

The course prediction value collection in one period moment of future that

counts for current time k backward;

is the following course set-point collection one to one constantly in

Diagonal angle, course weighting matrix W _ψBe unit matrix,

W _δBe diagonal angle, course weighting matrix, the value on its diagonal line gets 0.1～0.5,

I is the unit diagonal matrix, and subscript T is a transposition.

Above-mentioned phase of history does not have strict the qualification with following number constantly constantly, generally selects 1～50, comprises using the current time value to fill when the historical juncture, corresponding correlation did not have value.

Therefore, when the ship model identification with waves calibration is complete, the generalized predictive control law automatically updated with the ship model parameters, complete autopilot adaptive function; calibration parameters can be changed through the waves?

achieve the level of knowing the waves should function.

With reference to Fig. 4, the inventive method idiographic flow is following:

(1) measure boats and ships in the course of current time k ψ (k);

(2) make up ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), A (z ^-1) and B (z ^-1) be known multinomial coefficient, C (z ^-1) be unknown multinomial coefficient, this step with ship model as wave model solution wave model parameter C (z ^-1);

(3) with multinomial coefficient C (z ^-1) substitution ship model A (z ^-1) ψ (n)=B (z ^-1) δ (n-1)+C (z ^-1) e (n), n=k-n ₀..., k, 50≤n ₀≤200, find the solution uncorrelated random series e (n), calculate covariance

In conjunction with the classification of seas of the covariance of setting up in advance, confirm the course pipeline upper limit according to the classification of seas of current time k with the mapping relations demarcation current time k of classification of seas

And lower limit

Wherein, the mapping relations of covariance and classification of seas need to set up in advance, in actual sea situation, calculate covariance according to method of the present invention, statistical study its with the corresponding relation of meteorological department's observation classification of seas, thereby foundation mapping relations between the two.

The pipeline upper limit and lower limit

about the course set-point up and down symmetry and spacing between 10 °～60 °; High more according to classification of seas, the distance between the pipeline upper limit

and the lower limit

is carried out value more greatly.

The output controlled quentity controlled variable δ (k) of

(5) if current classification of seas is zero, according to A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k) calculating ∑ e ²(k), if ∑ e ²(k) greater than the predictive error threshold value, then get into step (6), otherwise, get into next and control constantly; Error threshold is an empirical value, chooses according to the control effect selection.

(6) will find the solution the multinomial coefficient C (z that obtains ^-1) substitution ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), this step is as model solution course, course model parameter A (z with ship model ^-1) and B (z ^-1); And then update controller, get into next and control constantly.Those skilled in the art will readily understand; The above is merely preferred embodiment of the present invention; Not in order to restriction the present invention, all any modifications of within spirit of the present invention and principle, being done, be equal to and replace and improvement etc., all should be included within protection scope of the present invention.

Claims (4)

1. Marine Autopilot self-adaptation control method is specially:

(1) measure boats and ships in the course of current time k ψ (k);

(2) make up ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), z ^-1Be unit factor retardation time, A (z ^-1) and B (z ^-1) be known multinomial coefficient, δ (k-1) is the controlled quentity controlled variable of moment k-1 ,-20＜e (k)＜20 finds the solution ship model and obtains multinomial coefficient C (z ^-1);

(3) with multinomial coefficient C (z ^-1) substitution ship model A (z ^-1) ψ (n)=B (z ^-1) δ (n-1)+C (z ^-1) e (n), n=k-n ₀..., k, 50≤n ₀≤200, find the solution uncorrelated random series e (n), calculate covariance

And then the classification of seas of the mapping relations demarcation current time k of the covariance of combination foundation in advance and classification of seas, confirm the course pipeline upper limit according to the classification of seas of current time k

And lower limit

(4) find the solution the output controlled quentity controlled variable δ (k) of controller

, wherein

$C=\left[\begin{array}{c}I\\ -I\\ C_{I/\mathrm{\&Delta;}}\\ -C_{I/\mathrm{\&Delta;}}\\ H\\ -H\end{array}\right],d_k=\left[\begin{array}{c}\mathrm{\&Delta;}\stackrel{\&OverBar;}{\mathrm{\&delta;}}\\ -\mathrm{\&Delta;}\underset{\&OverBar;}{\mathrm{\&delta;}}\\ \stackrel{\&OverBar;}{\mathrm{\&delta;}}-L{\mathrm{\&delta;}}_{k-1}\\ -\underset{\&OverBar;}{\mathrm{\&delta;}}+L{\mathrm{\&delta;}}_{k-1}\\ \stackrel{\&OverBar;}{\mathrm{\&psi;}}-\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}-Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}\\ -\underset{\&OverBar;}{\mathrm{\&psi;}}+\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}+Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}\end{array}\right],$

H＝C _D ^-1C _B，P＝C _D ^-1H _B，Q＝-C _D ^-1H _D，D(z ^-1)＝ΔA(z ^-1)，

C _B, H _BBe multinomial coefficient B (z ^-1) Toeplitz and Hankle matrix,

C _D, H _DBe multinomial coefficient D (z ^-1) Toeplitz and Hankle matrix,

C _{The I/ Δ}Be the Toeplitz matrix of polynomial expression I/ Δ, Δ=1-z ^-1,

is respectively steering wheel increment rudder angle upper physical limit and lower limit

is respectively the upper physical limit and the lower limit of steering wheel rudder angle amplitude

$f=-H^T{W}_{\mathrm{\&psi;}}({\underset{\&RightArrow;}{\mathrm{\&psi;}}}^{*}-\mathrm{P\&Delta;}\underset{\&LeftArrow;}{\mathrm{\&delta;}}-Q\underset{\&LeftArrow;}{\mathrm{\&psi;}}),$ S＝H ^TW _ψH+W _δ， $L=\left[\begin{array}{c}1\\ 1\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ 1\end{array}\right]$

plays the rudder angle value collection of one period continuous historical juncture of counting forward for a last moment k-1; plays the course value collection of the continuous historical juncture of section of counting forward for current time k;

equates with element number in

The rudder angle value collection in one period moment of future that counts for the current time k that finds the solution backward

The course prediction value collection in one period moment of future that

counts for current time k backward; is the following course set-point collection one to one constantly in

Diagonal angle, course weighting matrix W _ψBe unit matrix,

w _δBe diagonal angle, course weighting matrix, the value on its diagonal line gets 0.1～0.5,

I is the unit diagonal matrix, and subscript T is a transposition;

(5) if current classification of seas be zero and covariance

greater than the predictive error threshold value; Then get into step (6); Otherwise, get into next and control constantly;

(6) with multinomial coefficient C (z ^-1) substitution ship model A (z ^-1) ψ (k)=B (z ^-1) δ (k-1)+C (z ^-1) e (k), find the solution multinomial coefficient A (z ^-1) and B (z ^-1), and then update controller, get into next and control constantly.

2. Marine Autopilot self-adaptation control method according to claim 1 is characterized in that, said step (2) is found the solution ship model according to following mode and obtained multinomial coefficient C (z ^-1):

I. calculate ξ (k)=A (z ^-1) ψ (k)-B (z ^-1) δ (k-1)-(C ₁z ^-1+ C ₂z ^-2+ C ₃z ^-3+ C ₄z ^-4) e (k) renewal

Structure

Ii. ∑ being carried out singular value decomposes ${\stackrel{\&OverBar;}{U}}_k\left[\begin{array}{c}{\stackrel{\&OverBar;}{D}}_k\\ 0\end{array}\right]{\stackrel{\&OverBar;}{V}}_k^T=\mathrm{Svd}\left(\mathrm{\&Sigma;}\right)$ Confirm matrix Svd () is the svd function;

Iii. upgrade $U_k=U_{k-1}{\stackrel{\&OverBar;}{V}}_k^T,$ $D_k={{\stackrel{\&OverBar;}{D}}_k}^{-1};$

Iv. definition

Forms of characterization do ${\widehat{\mathrm{\&theta;}}}_k={[c_1,c_2,c_3,c_4]}^T,$ Upgrade

V. according to

evaluator coefficient

C(z ^-1)＝1+C ₁z ^-1+C ₂z ^-2+C ₃z ^-3+C ₄z ^-4。

3. Marine Autopilot self-adaptation control method according to claim 1 is characterized in that, said step (6) is found the solution ship model according to following mode and obtained multinomial coefficient A (z ^-1) and B (z ^-1):

1) Update

structure

2) ∑ being carried out singular value decomposes ${\stackrel{\&OverBar;}{U}}_k\left[\begin{array}{c}{\stackrel{\&OverBar;}{D}}_k\\ 0\end{array}\right]{\stackrel{\&OverBar;}{V}}_k^T=\mathrm{Svd}\left(\mathrm{\&Sigma;}\right)$ Confirm matrix

Svd () is the svd function;

3) upgrade $U_k=U_{k-1}{\stackrel{\&OverBar;}{V}}_k^T,$ $D_k={{\stackrel{\&OverBar;}{D}}_k}^{-1};$

4) definition Forms of characterization do ${\widehat{\mathrm{\&theta;}}}_k={[c_1,c_2,c_3,c_4]}^T,$ Upgrade

5) foundation

Evaluator coefficient A (z ^-1)=1+a ₁z ^-1+ a ₂z ^-2And B (z ^-1)=(b ₀+ b ₁z ^-1) z ^-1

4. Marine Autopilot self-adaptation control method according to claim 1; It is characterized in that; The said pipeline upper limit and lower limit

about the course set-point up and down symmetry and bound spacing between 10 °～60 °; Classification of seas is high more, and the spacing between the pipeline upper limit and the lower limit

is big more.

Images