CN103278152A - Fusion method of reference system for ship asynchronous position - Google Patents

Abstract

The invention relates to the field of power positioning information fusion, and specifically relates to a fusion method of a reference system for a ship asynchronous position used for data processing in a measurement system of a power positioning sensor The fusion method comprises the following steps of performing coordinate transformation and unit conversion on sensor measurement data to obtain coordinate values of a ship under a north-east coordinate system; filtering a subsystem; predicating o a predication is made in an N size; filtering in the N size; filtering on an i size; reconstructing wavelet; and carrying out fusion estimation in the N size. The fusion method is on the basis of observation model and measurement data on a plurality of scales, takes full advantages of complementary information on various scales according to a multi-scale distributed fusion theory, breaks through restriction of a single scale, reduces uncertainty, and forms relatively complete and consistent description of the system. At the same time, system and measurement information of an asynchronous multi-sensor are completely retained by the wavelet decomposition and reconstruction process, and defects of the asynchronous multi-sensor information treated by time rectification are overcome, so that the asynchronous multi-sensor has higher performance.

Description

The asynchronous position reference system fusion method of a kind of boats and ships

Technical field

The present invention relates to dynamically positioning information fusion field, be specifically related to the method that data are handled in a kind of dynamically positioning sensor measuring system.

Background technology

The safe operation of dynamic positioning of vessels system depend on vessel position and bow to uninterrupted measurement, in order to realize this goal, even under fault state, it is redundant that all measuring systems also keep.According to the Three Estate standard of International Maritime Organization (IMO) to the dynamic positioning system formulation, the dynamic positioning of vessels system must dispose position reference system and attitude of ship and the wind sensor of respective type and quantity, with guarantee that the contraposition of dynamically positioning boats and ships is put, the demand of attitude and environmental information.

The vessel position frame of reference can adopt the location survey sensor of different principle such as differential Global Positioning System, the underwater sound, side tension cords, laser, microwave, and the sensor of the most amounts of these multiple classes can be used in combination; Use the estimated accuracy that different types of multisensor can improve some observed quantity and description amount.

Usually there are situations such as sample frequency difference, observation is asynchronous, the sampling possibility is inhomogeneous in different accommodation sensors, and therefore the Multi-sensor Fusion algorithm research that carries out under asynchronous, many speed, the inhomogeneous sampling situation is very necessary.

The present invention is satisfying under the prerequisite that merges performance requirement, proposed the asynchronous position reference system fusion structure of boats and ships and the algorithm of a kind of high precision, high fault tolerance, high reliability, inventive result is significant to the performance, the realization hi-Fix that improve dynamic positioning system.

Summary of the invention

The objective of the invention is to propose a kind of have high precision, high fault tolerance, high reliability, can ensure the asynchronous position reference system fusion method of boats and ships of security, stability and the control accuracy of dynamic positioning system.

The object of the present invention is achieved like this:

The present invention includes following steps:

(1) the sensor measurement data obtain vessel position coordinate figure under the east northeast coordinate through coordinate transform, unit conversion;

(2) subsystem filtering: subsystem is carried out fault detect, if any fault, carry out fault-tolerant filtering; As non-fault, operation standard SRCKF filtering, completion status is estimated in the subsystem filtering;

(3) yardstick N goes up prediction: system state fusion is constantly estimated and corresponding evaluated error covariance matrix square root coefficient based on k, utilizes standard SRCKF filtering, obtains status predication value and predicated error covariance matrix square root coefficient;

(4) yardstick N goes up filtering: utilize yardstick N to go up measured value, carry out yardstick N according to the content of step (2) and go up subsystem filtering, obtain state estimation value and evaluated error covariance square root coefficient;

(5) the last filtering of yardstick i: 1) yardstick N is gone up information of forecasting and carry out wavelet decomposition, obtain the detail signal on the last state smoothing signal of yardstick i, the yardstick, and corresponding evaluated error covariance matrix; 2) go up measured value based on yardstick i, carry out yardstick i according to step (2) and go up filtering, smooth signal, corresponding evaluated error covariance matrix are measured renewal; Detail signal on the yardstick is not measured renewal;

(6) wavelet reconstruction: it is comprehensive that the detail signal on the smooth signal that upgraded on the yardstick i, corresponding evaluated error covariance matrix and the yardstick is carried out small echo to yardstick N, obtains yardstick N and go up system state goes up measured value based on yardstick i estimated value and corresponding evaluated error covariance matrix;

(7) yardstick N go up to merge estimates: yardstick N is gone up system state go up the estimated value of measured value and yardstick N based on yardstick N and go up system state and merge based on the estimated value that yardstick i goes up measured value, the system that obtains is engraved in when k+1 that the state based on global information merges estimated value on the yardstick N.

Beneficial effect of the present invention is: based on the observation model on a plurality of yardsticks and measurement data, according to multiple dimensioned distributed blending theory, take full advantage of the complementary information on each yardstick, break through the restriction of single yardstick, reduce uncertainty, form the description to the complete relatively unanimity of system.Because wavelet decomposition and restructuring procedure have intactly kept system and the metrical information of asynchronous multiple sensors, overcome the defective of being handled asynchronous multiple sensors information by temporal registration, thereby had higher performance simultaneously.

Description of drawings

Fig. 1 is asynchronous position reference system fusion structure;

Fig. 2 is that asynchronous position reference system merges block diagram.

Embodiment

General thought of the present invention is:

The measurement data of each position reference system positioned resolve, comprise coordinate transform, the unit conversion of sensor measurement data; Subsystem on each yardstick is carried out filtering, comprise the fault detect of sensor subsystem and select suitable filtering algorithm to obtain state estimation on each yardstick; According to the quality of data of each position measuring system, the state estimation of subsystem wave filter and data process system control command are set up fault-tolerant criterion, realize the dynamic redundancy combination of foreign peoples's position reference system adaptively; Based on multiple dimensioned distributed Fusion Estimation Algorithm, asynchronous position reference system is carried out the perfect information optimum fusion, obtain system state at the thinnest yardstick N of the highest sample frequency correspondence and merge estimation, realize that the system state of foreign peoples's multisensor merges estimation.Be specially: each yardstick arranges fault-tolerant control, guarantees that boats and ships diverse location frame of reference enters and withdraw from the ship's position measurement emerging system adaptively; To the different sensor of every class sample frequency, carry out the multiple dimensioned distributed fusion of multisensor and estimate that the system state that obtains foreign peoples's multisensor merges to be estimated.According to state estimation and data process system control command on each sensor measurement quality of data, each yardstick, set up fault-tolerant criterion, realize the dynamic redundancy combination between the sensor adaptively, guarantee that boats and ships diverse location frame of reference enters and withdraw from the ship's position measurement emerging system adaptively, realizes the dynamic optimization of dynamic positioning system sensor resource.Based on multiple dimensioned distributed state blending algorithm, select self-adaptation SRCKF as the nonlinear system filtering algorithm, asynchronous position reference system is carried out the perfect information optimum fusion.Hypothesis has obtained system and be engraved on the yardstick N fusion value based on global information when k simultaneously.The main algorithm step is as follows:

(1) measurement data filtering is prepared

The sensor measurement data obtain vessel position coordinate figure under the east northeast coordinate through coordinate transform, unit conversion;

(2) subsystem filtering content

1) fault detect of using mutant type and tolerant fail algorithm based on the filtering residual error, calculate saltant fault detect function, if any fault, carry out fault-tolerant filtering; Otherwise continue normal filtering;

2) use gradation type fault detect and tolerant fail algorithm, calculate gradation type fault detect function, if any fault, carry out fault-tolerant filtering; Otherwise continue normal filtering;

3) by residual computations filtering algorithm transfer criterion, if select self-adaptation SRCKF, then calculate the self-adaptation adjustment factor, carry out auto adapted filtering; Otherwise system optimal ground operation standard SRCKF filtering.

(3) circulation step

1) yardstick N goes up prediction

System state fusion is constantly estimated and corresponding evaluated error covariance matrix square root coefficient based on k, based on self-adaptation SRCKF filtering, obtains status predication value and predicated error covariance matrix square root coefficient.

2) filtering on each yardstick

A. utilize yardstick N to go up measured value, carry out yardstick N according to the content of step (2) and go up subsystem filtering, obtain state estimation value and evaluated error covariance square root coefficient.

B. yardstick i goes up filtering

A) the last information of forecasting of yardstick N is carried out wavelet decomposition, obtain the detail signal on the last state smoothing signal of yardstick i, each yardstick, and corresponding evaluated error covariance matrix.

B) go up measured value based on yardstick i, carry out yardstick i according to the content of step (2) and go up filtering, the corresponding evaluated error covariance matrix of smooth signal is measured renewal; But the detail signal on each yardstick is not measured renewal.

3) wavelet reconstruction

Utilize wavelet reconstruction, it is comprehensive that detail signal on the smooth signal that upgraded on the yardstick i, corresponding evaluated error covariance matrix and each yardstick is carried out small echo to yardstick N, obtains yardstick N and go up system state goes up measured value based on yardstick i estimated value and corresponding evaluated error covariance matrix.

4) yardstick N goes up to merge and estimates

Yardstick N is gone up system state go up the estimated value of measured value and yardstick N based on yardstick N and go up system state and merge based on the estimated value that yardstick i goes up measured value, the system that obtains is engraved in when k+1 that the state based on global information merges estimated value on the yardstick N.

Below in conjunction with accompanying drawing technical scheme of the present invention is elaborated.

(1) be research object with dynamic positioning of vessels redundant position frame of reference, its sample frequency becomes 2 ^j(j is integer) be relation doubly.The motion state model of dynamic positioning vessel is:

$\left\{\begin{array}{c}\stackrel{\·}{x}=u\cos \mathrm{\ψ}-v\sin \mathrm{\ψ}+{\mathrm{\ω}}_1\\ \stackrel{\·}{y}=u\sin \mathrm{\ψ}+v\cos \mathrm{\ψ}+{\mathrm{\ω}}_2\\ \stackrel{\·}{\mathrm{\ψ}}=r+{\mathrm{\ω}}_3\\ \stackrel{\·}{u}={\mathrm{\ω}}_4\\ \stackrel{\·}{v}={\mathrm{\ω}}_5\\ \stackrel{\·}{r}={\mathrm{\ω}}_6\end{array}\right.---\left(1\right)$

Variable x wherein, y, ψ be respectively boats and ships under local east northeast coordinate system the east northeast coordinate and bow to, u, v, r represent corresponding speed respectively.ω _iThe system interference of expression dynamic system model is zero-mean white Gaussian noise independent of each other.

Can get the discretize system equation by system's differential equation (1):

x(k)＝f(x(k-1))+w(k-1) （2）

Wherein, system matrix is f ∈ R ^{6 * 6}, systematic procedure noise w (k-1) is the zero-mean white Gaussian noise, and its covariance is:

Q＝diag[0.2,0.2,0.002,0.002,0.001,0.0001]

Observation equation is:

z _i(k)=h _ix _i(k)+v _i(k),k≥0,i＝1,2,3 （3）

Wherein, i is number of sensors, observation vector z _i=[x, y, ψ, u, v, r] ^T, h _i=I _{6 * 6}, measure noise v _iBe zero-mean white Gaussian noise sequence, covariance is respectively:

R ₁＝diag[0.2,0.2,0.002,0.002,0.001,0.0001]，R ₂＝diag[4,4,0.001,0.01,0.01,0.01]，R ₃＝diag[6,6,0.01,0.01,0.01,0.01]

Three position reference system (position reference system, sampling period PRS) is set at 0.25 second respectively, 0.5 second and 1 second, and system's initial value is: x (0)=[10,20,10,1,1.5,0.1] ^TAnd p (0)=diag ([1,1,1,1.5,1.5,0.5]), produce one group of theoretical value as the system state actual value by system equation formula (2), be designated as data-0, produce three groups of data data-1, data-2 and data-3 by data-0 and observation equation (3), tentation data data-1 is the measured value of PRS-1, and then the measured value of PRS-2 can carry out interval sampling by data-2 and obtains, and is designated as data-4; In like manner, with 1/4th former sample frequency data data-3 is carried out obtaining the measured value of PRS-3 every three samplings, be designated as data-5; With data-1, data-4 and the data-5 measured value as three asynchronous-sampling PRS.

(2) choice criteria SRCKF of the present invention is as the nonlinear system filtering algorithm, and hypothesis has obtained system and be engraved on the yardstick N fusion value based on global information when k simultaneously.

(3) respectively organize the sensor measurement data through coordinate transform, unit conversion, obtain vessel position coordinate figure under the east northeast coordinate.

(4) subsystem filtering: subsystem is carried out fault detect, if any fault, carry out fault-tolerant filtering; Otherwise continue operation standard SRCKF filtering, completion status is estimated in each subsystem filtering.

(5) according to state estimation and data process system control command on the quality of data of each position measuring system, each yardstick, set up the fault-tolerant criterion of each dimension location measurement subsystem.

(6) yardstick N goes up prediction, and system state fusion is constantly estimated and corresponding evaluated error covariance matrix square root coefficient based on k, utilizes standard SRCKF, obtains status predication value and predicated error covariance matrix square root coefficient.

(7) yardstick N goes up filtering, utilizes yardstick N to go up measured value, carries out yardstick N according to the content of (4) and goes up subsystem filtering, obtains state estimation value and evaluated error covariance square root coefficient.

(8) yardstick N goes up prediction, and system state fusion is constantly estimated and corresponding evaluated error covariance matrix square root coefficient based on k, utilizes standard SRCKF, obtains status predication value and predicated error covariance matrix square root coefficient.

(9) yardstick i goes up filtering, uses the Daubechies4 wavelet filter and carries out wavelet decomposition and reconstruct:

A) the last information of forecasting of yardstick N is carried out wavelet decomposition, obtain the detail signal on the last state smoothing signal of yardstick i, each yardstick, and corresponding evaluated error covariance matrix.

B) go up measured value based on yardstick i, carry out yardstick i according to the content of (4) and go up filtering, smooth signal, corresponding evaluated error covariance matrix are measured renewal; But the detail signal on each yardstick is not measured renewal.

(10) utilize wavelet reconstruction, it is comprehensive that detail signal on the smooth signal that upgraded on the yardstick i, corresponding evaluated error covariance matrix and each yardstick is carried out small echo to yardstick N, obtains yardstick N and go up system state goes up measured value based on yardstick i estimated value and corresponding evaluated error covariance matrix.

(11) yardstick N goes up to merge and estimates, yardstick N is gone up system state go up the estimated value of measured value and yardstick N based on yardstick N and go up system state and merge based on the estimated value that yardstick i goes up measured value, the system that obtains is engraved in when k+1 that the state based on global information merges estimated value on the yardstick N.

So far, finished the information fusion of dynamic positioning of vessels foreign peoples's position reference system.

Claims (1)

1. the asynchronous position reference system fusion method of boats and ships is characterized in that, comprises the steps:

(1) the sensor measurement data obtain vessel position coordinate figure under the east northeast coordinate through coordinate transform, unit conversion;

(2) subsystem filtering: subsystem is carried out fault detect, if any fault, carry out fault-tolerant filtering; As non-fault, operation standard SRCKF filtering, completion status is estimated in the subsystem filtering;

(3) yardstick N goes up prediction: system state fusion is constantly estimated and corresponding evaluated error covariance matrix square root coefficient based on k, utilizes standard SRCKF filtering, obtains status predication value and predicated error covariance matrix square root coefficient;

(4) yardstick N goes up filtering: utilize yardstick N to go up measured value, carry out yardstick N according to the content of step (2) and go up subsystem filtering, obtain state estimation value and evaluated error covariance square root coefficient;

(5) the last filtering of yardstick i: 1) yardstick N is gone up information of forecasting and carry out wavelet decomposition, obtain the detail signal on the last state smoothing signal of yardstick i, the yardstick, and corresponding evaluated error covariance matrix; 2) go up measured value based on yardstick i, carry out yardstick i according to step (2) and go up filtering, smooth signal, corresponding evaluated error covariance matrix are measured renewal; Detail signal on the yardstick is not measured renewal;

(6) wavelet reconstruction: it is comprehensive that the detail signal on the smooth signal that upgraded on the yardstick i, corresponding evaluated error covariance matrix and the yardstick is carried out small echo to yardstick N, obtains yardstick N and go up system state goes up measured value based on yardstick i estimated value and corresponding evaluated error covariance matrix;

(7) yardstick N go up to merge estimates: yardstick N is gone up system state go up the estimated value of measured value and yardstick N based on yardstick N and go up system state and merge based on the estimated value that yardstick i goes up measured value, the system that obtains is engraved in when k+1 that the state based on global information merges estimated value on the yardstick N.

Images