JPS59179497A - Automatic steering device - Google Patents

Abstract

PURPOSE:To enable an automatic steering device to perform optimum weather regulation, by a method wherein, based on the detected value of the yawing motion of a ship hull, a parameter under a marine weather condition is determined, and a noise is estimated so that large or small value of the observation noise of a karman filter corresponds to large or small value of the parameter respectively. CONSTITUTION:An azimuth deviation between a set azimuth psi1 and a bow azimuth psi of a compass 4 is detected by a comparator 8, and the result is fed to a karman filter part 9 togetherwith the furning angular velocity psi' of an angular velocity meter 5 and the steering angle delta of a steering angle indicator 7. The karman filter part 9 finds optimum estimated values psie'' and psi'' of an azimuth deviation psie and furning angular velocity psi' on a basis of these data, a ship speed S, and the marine weather parameter of a parameter computing part 15 to send the estimated values to a regulating part 3. The regulating part 3 finds an instruction steering angle delta* based on these estimated values and aforesaid azimuth deviation psie, and based on the instruction steering angle, a steering device 2 is steered.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an automatic steering system, and more particularly to an automatic steering system in which weather adjustment is performed using a Kalman filter.

The autopilot mechanism of ships has traditionally been equipped with a weather adjustment device. This weather adjustment device is designed to eliminate wasteful rudder in response to high-frequency components of turning motion induced by waves when sea conditions become rough, and to prevent loss of horsepower and wear and tear on the rudder, etc. be. However, conventional weather adjustment devices that use nonlinear elements such as linear filters, butterflies, dead zones, and double steering angle ratios are not necessarily sufficient for the purpose of removing only high-frequency components that cannot be suppressed by steering. Since components related to steering (including components caused by wind that should originally be controlled) are also removed at the same time, the lack of coordination in the steering control system has the disadvantage of limiting performance. Furthermore, since course-keeping performance is greatly affected by how the adjustment values are set, it is virtually impossible to obtain the optimum 71-degree adjustment, for example, when the navigator manually adjusts the A. .

In order to improve the conventional weather adjustment that has such drawbacks, the present inventor developed a Kalman filter (optimal filter).
We devised a weather adjustment method using The Kalman filter is a filter that provides a computer-friendly algorithm for obtaining highly reliable optimal estimates by fully applying the least squares method to the estimated values obtained from each of the system equations and observation equations, and is effective for disturbance processing. be. By filtering the azimuth deviation (time-series signal) using this Kalman filter, it is possible to remove only the noise component, that is, the high frequency component that cannot be suppressed by steering. The Kalman filter method will be explained below.

First, we will discuss the system equation and observation equation that are the basis of the Kalman filter. The system equation is the maneuvering response of the ship, and according to Nomoto's first-order approximation model, T≠+φ=K(δ+V)...Mong, Mong, Mong, Tong, (1
) where K, T: Maneuverability index φ: Turning angular velocity δ: Rudder angle V Expressed by nibrant noise. The constant difference equation that discretizes this is 1 (k
+1)=AZ(k)+Bu(k)+rv(
k)・--Song (2) Here, F=Bu(k)-δ(k) T T v(k) Nibrandt noise, normal white noise (mean value 07 variance Q) τ: Sampling time .

Next, the observation equation is discretized and expressed as Y (k+1
) = Cχ (k edition 3) + W (J 4-track - (3) where, w(k): observation noise, normal white noise (average value 0.
covariance R).

In contrast to (2) and (3), the discrete-time Kalman filter is expressed by the following equations (4) to (8) as well known.

z (k+1 + k)=Az (k 1k
) 10f3u (k) ---(4)P(k+11k)
-AP(klk)AT'+FQ〆・・・・・・・・・
... (5) G (+1 to ') = P (k + 1 day OCT [C
P(k+1 eyes XCT ten ding 1 ・・・・・・・・・・・・・・・・・・・・・・・・
・・・・・・・・・ (6)9 (k+1 l k+1
)=”'!, (k+1 l k) +G(k+1)
X[Y(k+1)-Cte(k+11k)]...
・・・－・・・・・・・・・・・・・・・・・・・・・
... (force P (+1 to k+11) = [I-G (k+1)
C]P(k+11k)・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・ (8) Here, i (k+11k) is the estimated value of χ at the time (k+1) when the observation data at the time point was obtained. represents. Further, '' is the optimal estimated value, and P and G are the covariance matrix of the error I and the Hen vector of the filter, respectively.

! , ta represents transposition.

If formulas (4) to (8) are calculated every time data is sampled, the optimal estimate will be Z (k+1 + k+1
) is obtained. In the system considered here (the hull control system and its observation system), the dynamics of the signal process and measurement process do not change over time, so the system can be expressed as an equation with constant coefficients, and changes in disturbances such as waves. is also gradual, so if we consider a short period of about 30 minutes, its statistical properties can be treated as stationary. Therefore, the Kalman filter becomes a linear dynamic stem with constant coefficients. Specifically, the error covariance matrix and the Kalman filter gain matrix are constant.

The steady state error covariance and filter gain are (5)
, (6), From formula (8), Q=[mi+Fqf)c
T(c(AI5f+rcg")C"+J(:]'...
・・−・・−・・−・・・・・・・
・・・・・・−・(9)P=CI−(APAT+r
QrT) cT(CCAE'AT+ rQr”)XC
T+R 1 x C] [Al5AT+rQF1]...
・・・・・・・・・・・・・・・・・・・・・・・・・・・
(10) Here, G and P represent estimated values of filter gain and error covariance in a steady state, respectively.

” ff0) formula is Newton-Raphso
It can be solved using the n method, and G can be obtained by substituting the result into equation (9). Therefore, in actual signal processing, it is only necessary to fully calculate equations (4) and ('t) in real time, and implementation is easy.

Here, the average value of plant noise is set to O, but in reality, there is no plant noise with an average value of 0, and it always has a certain bias component. Therefore, instead of formula (2), χ (k + 1) = = Az (k) + Bu (k)
+1' (k) +va)・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・ (1]) Here, ■. Although it is possible to use the average value of the nibrant noise, this would result in an increase in the amount of calculation.

Therefore, estimation is performed separately from the Kalman filter.

Bias component of this plant noise■. Since the system is a closed loop and the bias component of the disturbance applied to the hull and the average value of the manipulated variable (rudder angle) are considered to be balanced in a steady state, the average value of the azimuth deviation signal is used. ,・・・・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・・・・・・・
... (12) can be estimated (actually, a redundant difference equation is used).

This is basically the same as the integral control performed in conventional autopilots. However, h is an arbitrary constant and is determined in consideration of the stability of the system.

On the other hand, the variance Q of plant noise and the covariance R of observation noise are so important that the quality of filtering is determined by how appropriately they are estimated. In particular, plant noise is difficult to grasp quantitatively and poses the greatest problem when applying a Kalman filter to an actual system. First, since it is thought that normal observation systems are not affected much by changes in oceanographic conditions, the covariance of observation noise is treated as constant. On the other hand, Q is Godbo's extension of Mehra's online estimation of white Gaussian noise covariance
It is conceivable to obtain it using the method of le. I will omit the details, but it is a variance estimation algorithm that can be applied even when plant noise and observation noise are correlated and the average value of the noise is not zero. This is a very useful method in the sense that it can be understood quantitatively.

However, by using an algorithm that calculates this Q using the Qodbole method, we can obtain the optimal estimated values of the heading deviation (set heading - heading) and turning angular velocity, and calculate the steering gear P engineering (PI
D) It was found that the following problems occur when the control is performed. That is, in Ca1m sea, the steering angle change was smooth, but course keeping was poor. On the contrary, Rough se
In case a, the rudder angle suddenly increased, causing overuse of the rudder gear. This is because Godbole's method estimates the variance of the plant noise too small when the sea is Ca1m, and the filtering effect was achieved, but the optimal estimate was delayed and the yawing became large.
In the case of a, the value of Q becomes very large because high frequency components (close to the rolling period) included in the azimuth deviation or turning angular velocity are considered to be caused by plant noise. This is because the high frequency components included in the angular velocity will be faithfully followed.

Furthermore, the stability of the estimated value itself was poor, resulting in large variations in the 9 values under the same sea state.

Considering a time-series signal, if we consider that the signal is a combination of the original meaningful signal component and the meaningless noise component, filtering is the process of separating the noise component as much as possible.

The goal is to extract only meaningful signal components.

Applying this to the present case, the signal components that originally have meaning are the low frequency components close to the steering system frequency included in the azimuth deviation and turning angular velocity, and the meaningless noise components. It's not just the error of the gyro, etc., but it can be suppressed by steering induced by waves.
This includes high-frequency components that are not present in the air. Taking this into account, if we look at system equation (3) again, we get
The system equation itself does not express any such clear distinction. When the Q□dbole algorithm is applied, if the sea conditions are rough and there are many high-frequency components that cannot be suppressed by steering, they are all considered to be induced by plant noise, and the estimated value of the variance Q of the plant noise is extremely low. becomes a large value. In this way, when the estimated value of Q becomes large, all high-frequency components (( This is because the system equation does not sufficiently express the filtering that we want, but there are many kinematic problems in expressing only the high frequency components induced by waves, and the system equation There is also the disadvantage that the order of the Kalman filter increases and the amount of calculation increases.

Godbole's method 1d assumes white noise and uses the autocorrelation of the invasion process, so if the data sampling time is short and the autocorrelation of the invasion process is strong, the variance cannot be calculated, and the number of data is If it is too small, the estimation value will be more irregular, and therefore the sampling time must be longer and the amount of data must be sufficiently large, which has the disadvantage of requiring a long time to prepare for calculation.

The inventor of the present invention has conducted extensive research in view of the drawbacks of the above-mentioned method, and has found that by devising a method for estimating filter coefficients, it is possible to prevent the optimal estimated value of azimuth deviation, etc. from being influenced by high frequency components caused by waves. This discovery led to this invention.

The present invention is to fully utilize the functions of a 3-relational filter in a simple configuration to improve the stability and coordination of automatic steering system control, and to automatically and reliably perform optimal weather adjustment. The purpose is to provide an automatic steering system that can.

The present invention detects turning movements such as the heading, turning angular velocity, etc., determines the degree of waves, that is, the parameter of the sea state based on this detected value, and the relative value of the observation noise of the Kalman filter is determined depending on the magnitude of this parameter. The above objective is achieved by estimating the noise so that the following occurs and applying a limiter to the noise when the sea is calm.

First, the principle of an embodiment of the present invention will be explained.

The Kalman filter algorithm using the Qodbole method described above assumes that the observation system will not be affected by rough sea conditions. As long as we consider this, it is necessary to express the high frequency components induced by waves in the system equation. Therefore, in this embodiment, the system equation (2) is used, and although the hull is actually driven by waves and a high frequency component is generated, it was decided to include this in the observation system and process it. That is, assuming that when the sea conditions become rough, the side threads will also be affected and the observation noise will increase, and a rough estimate will be made from the parameters of the sea conditions. Due to the nature of the Kalman filter, the general size and the filtering effect size match, so (Q is assumed to be constant for simplicity), when the sea condition becomes rough and 1 becomes large, high frequency components induced by waves are filtered out. This is what VC does.

At this time, since the optimal estimated value is obtained mainly from the system equation, it is not filtered by the high frequency component 1 close to the natural frequency of the steering system.

On the other hand, if sea conditions are expressed as Beaufort class parameters, interference from individual differences can be avoided, and the influence of sea conditions on the hull control system is not proportional to the Beaufort class, but varies depending on the direction of waves and swells. Depends on it.

As is well known, even if the Beaufort class is the same, if the direction of the swell is diagonally to the rear, it will have a large effect on the control system, and if it is in the front, it will have little effect. Therefore, as the parameters of the sea state seen from the autopilot, parameters obtained by detecting the turning motion of the ship, such as turning angular velocity, and performing predetermined calculation processing are used. For example, if the turning angular velocity r, ITL S, and e are parameters of the sea state that are not affected by the controller, and R is estimated in proportion to this parameter, the plant noise and observation noise are Q2 const...・・・・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・・・・・・・ (1
3). Here, m and In1d are positive arbitrary constants,
It is determined experimentally, taking into consideration the stability of the system.

However, at Ca1m sea, the level of the turning angular velocity induced by waves is comparable to the level induced by the steering angle, and the turning angular velocity r, m, s.

The influence of the controller becomes large in is constant. On the other hand, ](o
In ugb sea, the covariance R of the observation noise was increased in proportion to the turning angular velocity r, turtle S, and E, so if there is a delay in the optimal estimate, R may be kept constant as necessary (see Figure 3). Next, a detailed explanation will be given according to FIG.

FIG. 1 is a system diagram of an automatic steering system using a Kalman filter for weather adjustment. In the figure, reference numeral 1 denotes a hull as a controlled object, and by steering a rudder 2 installed at the stern of the hull I with an adjustment section 3, course keeping control of the target course is performed. There is. The hull l is equipped with a compass 4 for measuring the heading ψ, turning angular velocity, ship speed S, and rudder angle δ, respectively. Angular velocity meter5. A speed log 6 and a steering angle indicator 7 are provided. After the comparator 8 detects the azimuth deviation ψ6 (-ψ1-ψ) between the set azimuth ψ1 set by a course setting device (not shown) and the boat light azimuth ψ output from the compass 4, the turning angular velocity is determined and the rudder angle δ is determined. The signal is sent to the Kalman filter unit 9. This Kalman filter section 9 includes: an R calculation section 10 that estimates observation noise, a data generation section 11 that outputs a Q value of plant noise, and a hull steering index that estimates T from the ship speed S to obtain filter coefficients A and B. A, B calculating section 12, P, Q calculating section 13 which calculates error covariance and filter gains P, G, azimuth deviation ψ. and an optimum value calculation unit 14 that calculates ψ from among the optimum estimated values of the turning angular velocity. The angular velocity meter 5 has a function as a turning motion detecting means for the hull l, and the detected turning angular velocity is sent to the parameter calculating section 15. This parameter calculation section 15
calculates r, m, and s based on data within a certain period of time and sends them to the R calculation unit 10 as sea parameters. This R calculation section 1
0 is the observation noise R according to the diagram in Figure 3? Do all the estimations. When the ship speed S changes beyond a certain level, the A and B calculation units 12
Hull maneuverability index K corresponding to the ship speed S.

It has the function of obtaining T. This is to deal with fluctuations in system characteristics, and actually the filter coefficients A and B in equation (2) are calculated. The P, G calculation section 13
, Q, A, B data iC basis @QO) +(
9) Estimate l' and Gi using equations. Kalman filter section 9
is initialized when the observation noise R can be estimated or when the ship speed S changes beyond a certain level, and when the optimum value calculation section 15 is initialized, the initial optimum estimated value χ (010) and appropriate values as the steering angle u(0), and A and B are sent from the A and B calculation unit 12.
ψ as Y data sent from the comparator 8 and the angular velocity meter 5 based on the G data sent from the data and the P, G calculation section 13. ,? ) and each time the rudder angle δ is sampled, the optimal estimated value χ is sequentially calculated according to equation (4) l (7)
(k+11 ic1i ) is calculated.

The optimal estimated value φ is thus filtered by the Kalman filter unit 9. , φ is the PD adjustment section 1 in the adjustment section 3
6, and the calculation of □δ,' -KF (φ8+T=ゐ) is performed. On the other hand, in the adjustment section 17 in the adjustment section 3, δ;=LΣψ is obtained by converting the equation (12) into a constant difference. (k) An operation of 1゛l is performed. These bl l 6M are added by an adder 18 and output to the steering gear 2 as a command rudder angle δ°.

However, as for the bias component of the plant noise, instead of processing it in the adjusting section, a bias processing section 20 is installed between the steering angle indicator 7 and the optimum value calculation section 14 of the Kalman filter, as shown in FIG. , this bias processing section 2
It is also possible to perform the calculation of 6=6-〒12゛ψ·(k) at 0, and use the steering angle δ from which the bias component is removed as the steering angle input of the total Kalman filter.

Next, as a second embodiment, instead of formulas (13) + (14), for example,
5) R=const, ・・・・・・・・・・・・
・・・・・・・・・・・・・・・・・・・・・・・・
・(1sn is a positive constant and is determined experimentally.

In addition, the plant noise Qi may be estimated to be inversely proportional to the parameters of the sea state. This is because, due to the nature of the Kalman filter, the magnitude of the filtering effect matches the magnitude of Q (1), and if Q becomes small due to rough sea conditions, high frequency components induced by waves will be filtered out. However, Ca+1m s
ea and low sea, a limiter is provided in the same way as in Fig. 3 (see Fig. 4).

Incidentally, the turning motion of the ship may be detected from the bow direction, and parameterization to the sea state may be performed using r, m, s, or an absolute value average isogeometric method. Estimation was also done using methods other than proportional/inverse proportional, and the results were good. In the second embodiment, each Q and R are CON S T,
It is not limited in any way.

As explained above, according to the present invention, it is possible to stably estimate the noise coefficient 'i of the Kalman filter using a simple algorithm, reduce the burden on the computer, and
It is possible to suppress high frequencies induced by waves by the steering system, without causing wasteful steering due to following high frequency components caused by waves at 1 m sea, or causing a delay in the optimal estimated value and causing the ship to meander significantly at rough seas. Only the components can be reliably filtered out, so an automatic steering machine with excellent weather adjustment performance and extremely high course keeping performance can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a system diagram of an automatic steering system according to an embodiment of the present invention,
FIG. 2 is a block diagram showing another embodiment of the plant noise bias processing method, FIG. 3 is a diagram showing the observation noise estimation method in the first embodiment, and FIG. 4 is a block diagram showing the plant noise bias processing method in the second embodiment. 3 is a diagram showing a method for estimating noise Q. FIG. ■... Hull, 5... Angular velocity meter, 9... Kalman filter section, 10... R calculation section, 11... Data generation section, 15... Parameter calculation section. Patent Applicant: Tokyo Co., Ltd. Meter Figure 2 φ % Te Figure 3 Mustsho Kinka

Claims (1)

[Claims] (1) In an automatic steering system equipped with a Kalman filter for weather adjustment, a device for detecting the canthus movement of the ship, and a device for calculating parameters of sea conditions by performing predetermined arithmetic processing on the detected data of the canthus movement. , and the noise is estimated so that the relative value of the observation noise to the plant noise of the -) Norman filter becomes large or small depending on the size of this parameter, and the noise V limiter is applied during calm sea 111c. A self-a steering system featuring: