JPH07306251A - Target motion analyzing method - Google Patents

Abstract

PURPOSE:To determine an optimal timing for starting the motion analysis of a target. CONSTITUTION:An orientative information calculating section 12 receives a signal IN and calculates a direction observation amount beta (t) and an error standard deviation sigma(t) An initial value calculating section 14A receives the observation amount 4 (t) and determines an initial value #(vX0) of motion analysis which is then delivered to an information matrix calculating section 15 and an inner state amount estimating section 18A. The calculating section 15 receives the direction observation amount beta(t), the error standard deviation sigma(t), an acceleration signal and the initial value #(vX0) and calculates an estimated information matrix #F. An eigen value calculating section 16 receives the estimated information matrix #F and calculates an eigen value thereof. An observability deciding section 17 delivers an observability decision signal to the estimating section 18A based on the eigen value. The estimating section 18A delivers an estimated values #{vx2 (t)}, #{vx*2(t)} of the inner state amount of a target, determines based on the observability decision signal, from an output terminal OUT.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention receives a sound radiated from a target by a receiver sensor array attached to a moving body such as a ship that can move, and moves it from an observation amount disturbed by noise. The present invention relates to a target motion analysis method for estimating an internal state quantity relating to the position and speed of a target being operated.

[0002]

2. Description of the Related Art FIG. 2 is a geometrical explanatory view showing an observation system and a motion system in a conventional target motion analysis method. In FIG. 2, (X, Y) is a fixed coordinate system of the origin O, 1 is a moving body, and 2 is a target. vx ₁ (t) (where v means a vector) is the position vector of the vehicle 1 at time t, and vx ₂ (t) is the position vector of the target 2 at time t. Further, r (t) is the distance between the vehicle 1 and the target 2 at time t ‖vx ₂ (t) -vx
₁ (t) ‖ (where ‖ ‖ represents the norm of the vector), θ (t) is the azimuth angle of the target 2 with respect to the Y axis as seen from the navigation vehicle 1 at time t. The target motion analysis is
The target position and speed are identified from the observation time series of the azimuth angle and frequency of the target sound source disturbed by noise, assuming that the target is performing uniform linear motion.

Prior to the description of the prior art, the principle of operation of the desired motion analysis method will be described. Goal 2 at time t
Azimuth angle β (t), which is the measurement result of the azimuth angle θ (t) of
Is expressed by the following equation (1). β (t) = θ (t) + n (t) (1) where n (t); observation noise whose mean value at time t is 0 and whose variance at time t is σ ² (t). Next, the relative position vector vx (t) viewed from the observer with respect to the target 2 is expressed by the following equation (2). _{vx (t) = vx 2 (} t) -vx 1 (t) = [r x (t) r y (t)] T ··· (2) where, T; internal state vX in transposed time t (t ) Is expressed by the following equation (3).

[Equation 1] However, *; the first-order derivative is the internal state quantity at time t = t ₀ is vX ₀ = vX
(T ₀ ), at time t = t ₁ , ..., T _n , the evaluation function L (vX _x of the internal state quantity vX ₀ is calculated from the set of azimuth observation quantities β (t) expressed in time series. ₀ ) to the next (4)
Expressed as an expression.

[0004]

[Equation 2] Internal state quantity vX that minimizes this evaluation function L (vX ₀ ).
By calculating ₀ , the internal state quantity v can be calculated by the following equation (5).
X (t) is estimated and the optimum value of this internal state quantity vX (t) is obtained.

[Equation 3] However, in order to obtain the internal state quantity vX ₀ that minimizes the **; second derivative evaluation function L (vX ₀ ), it is necessary to solve the non-linear equation represented by the following equation (6). ∂L (vX ₀ ) / ∂X ₀ = 0 (6) In this case, the observation noise n (t) is set to 0 in the equation (1), and the pseudo-linear equation expressed by the following equation (7) is obtained. Solve,
The initial value of motion analysis # (vX ₀ ) is determined.

[0005]

[Equation 4] Next, a conventional desired motion analysis method will be described. FIG. 3 is a functional block diagram of a desired motion analysis device for implementing a conventional desired motion analysis method. This target motion analysis device
It has an input terminal IN1 for inputting a signal in from a wave receiver sensor array (not shown) attached to the vehicle 1 in FIG. The input terminal IN1 is connected to the input side of the azimuth information calculation unit 12. The azimuth information calculation unit 12 inputs the signal in, and is the azimuth observation amount β (t) and the azimuth observation amount β (t) which are the measurement results of the azimuth angle θ (t) of the target 2 in FIG.
Is a means for calculating the error standard deviation σ (t) of the above in a time series manner. Further, the target motion analysis device is provided with a running body maneuver determination unit 13 which receives the acceleration signal α of the running body 1 input from the acceleration signal input terminal IN2 and generates a control signal S13 for motion analysis. . On the other hand, the azimuth information calculation unit 12
On the output side of the azimuth observation amount β (t) and the initial value of motion analysis from the control signal S13 of the navigation vehicle maneuver determination unit 13
It is connected to the input side of the initial value calculation unit 14 for calculating (vX ₀ ). The output side of the initial value calculation unit 14 is connected to the input side of the internal state quantity estimation unit 18. The internal state quantity estimating unit 18 determines the error standard deviation σ of the initial value # (vX ₀ ) of the motion analysis, the azimuth observation amount β (t), and the azimuth observation amount β (t).
By inputting (t), the target 2 which is the internal state quantity of the target motion.
Of the position vector vx ₂ (t) of
₂ (t)}, and target 2 velocity vector vx ^* ₂ (t)
Is a means for calculating an estimated value # {vx ^* ₂ (t)} of
The output side of the internal state quantity estimation unit 17 has an estimated value # {vx
₂ (t)}, # {vx ^* ₂ (t)} are connected to the output terminal OUT which outputs the target motion analysis result.

Next, a conventional desired motion analysis method using this desired motion analysis device will be described. Acceleration signal input terminal I
The acceleration signal α of the navigation vehicle 1 in FIG. 2 input from N2
Is input to the vehicle maneuver determination unit 13. In the navigation vehicle maneuver determination unit 13, when the integrated value S of the acceleration signal α represented by the following equation (8) exceeds a preset threshold value, the navigation vehicle 1 changes its needle (hereinafter referred to as a maneuver). The control signal S13 is sent to the initial value calculation unit 14 so as to start the motion analysis.

[Equation 5] On the other hand, the signal in from the signal source received by the receiver sensor array is input to the input terminal IN1, and the azimuth information calculation unit 1
Entered in 2. When the signal in is input to the azimuth information calculation unit 12, the azimuth observation amount β (t) and the azimuth observation amount β (t).
The error standard deviation σ (t) of is calculated. Initial value calculation unit 14
Then, n sets of azimuth observation amounts β (t) at t = t ₁ , ..., T _n are input, and the maneuvering vehicle maneuver determination unit 1
When the control signal S13 is input from 3, the initial value # (vX ₀ ) of the motion analysis is obtained based on the equation (7), and the result is output to the internal state quantity estimating unit 18. Internal state quantity estimation unit 1
Eighth is the estimated value of the position vector vx ₂ (t) of the target 2 which is the internal state quantity of the target motion, by solving the non-linear equation (6) using the initial state quantity # (vX ₀ ) as the initial value of the motion analysis.
{Vx ₂ (t)} and velocity vector vx ^* _{2 of} target ₂
An estimated value # {vx ^* ₂ (t)} of (t) is obtained, and the result is output from the output terminal out.

[0007]

However, the conventional target motion analysis method has the following problems. That is, since the observability determination indicating whether or not to start the motion analysis is performed based on the determination of the maneuver of the navigation body 1, if the accuracy of the bearing observation amount β (t) is insufficient, The observability judgment becomes inaccurate. Therefore, there is a case that the optimum timing for starting the motion analysis of the target 2 may not be obtained, the limited calculation resources cannot be effectively utilized, and there is a problem that the operator is given a meaningless analysis result.

[0008]

According to the first aspect of the invention, in order to solve the above-mentioned problems, the sound emitted from a target is received by a receiver sensor array attached to a movable vehicle and the target is received. Azimuth angle is measured, the azimuth observation amount which is the measurement result of the azimuth angle and azimuth information calculation processing for calculating the accuracy of the azimuth observation amount in time series, the azimuth observation amount, the accuracy of the azimuth observation amount,
And an initial value calculation process for calculating an initial state quantity for solving a target motion analysis by solving a pseudo-linear equation obtained by setting the observation noise component of the azimuth observation amount to 0 from the integrated value of the acceleration of the vehicle Using the amount as the initial value of the target motion analysis, an internal state quantity estimation process for obtaining an estimated value of the target state quantity based on the time series of the azimuth observation quantity and the time series of the accuracy of the azimuth observation quantity is sequentially performed. The following measures are taken in the target motion analysis method. That is, based on the azimuth observation amount, the accuracy of the azimuth observation amount, the integrated value of the acceleration of the vehicle and the initial state amount, an estimated information matrix that is an inverse matrix of the estimation error covariance matrix of the target state amount is calculated. Information matrix calculation processing, eigenvalue calculation processing for calculating eigenvalues of the estimated information matrix, and observability determination that generates an observability determination signal that is a control signal for starting the motion analysis of the target based on the eigenvalues. Processing and the internal state quantity estimation processing based on the observability determination signal. In a second invention, in the desired motion analysis method of the first invention, it is determined whether or not to start the desired motion analysis based on the eigenvalue obtained by singular value decomposition of the estimated information matrix.
In the third invention, in the target motion analysis method of the first invention, only the maximum eigenvalue and the minimum eigenvalue are obtained to determine whether to start the target motion analysis. Fourth
In the invention of claim 1, in the target motion analysis method of the first invention,
The distribution range of eigenvalues is estimated, and it is determined whether or not to start the target motion analysis based on the estimated distribution range.

[0009]

According to the first aspect of the invention, since the target motion analysis method is configured as described above, in the information matrix calculation process, the target direction observation amount calculated in the direction information calculation process and the accuracy of the direction observation amount. An estimated information matrix, which is the inverse of the estimation error covariance matrix of the target state quantity, is calculated based on the integrated value of the acceleration of the vehicle and the initial state quantity calculated in the initial value calculation process. Next, in the eigenvalue calculation process, the eigenvalue of the estimated information matrix is calculated. Further, in the observability determination process, an observability determination signal which is a control signal for starting the motion analysis of the target is generated based on the eigenvalue.
Internal state quantity estimation processing is performed based on this observability determination signal. According to the second invention, in the eigenvalue calculation process of the first invention, the eigenvalue is obtained by performing singular value decomposition on the estimated information matrix. In the observability determination process, it is determined whether to start the target motion analysis based on the eigenvalue. According to the third aspect, in the observability determination process of the first aspect, it is determined whether or not to start the target motion analysis by obtaining only the maximum eigenvalue and the minimum eigenvalue of the estimated information matrix. According to the fourth aspect, in the observability determination process of the first aspect, it is determined whether to start the target motion analysis based on the estimated eigenvalue distribution range. Therefore, the above problem can be solved.

[0010]

First, the principle of a desired motion analysis method, which is the basis of the embodiment of the present invention, will be described. Time t = t ₁ , ...
The observability of the target motion analysis at t _n is determined by the information matrix F shown in the following equation (9). If F is regular, it is observable, and if F is not regular, it is non-observable. .

[Equation 6] Therefore, the information matrix F is a function of the variance σ ² (t) of the observation noise of the azimuth observation amount β (t), and the azimuth observation amount β (t ) Observability depends on the angle measurement accuracy. That is, when the angle measurement error is relatively small, it becomes observable even if the integrated value of the acceleration component α of the navigation body 1 is small, but when the angle measurement error is relatively large, it becomes observable. The lower limit value of the integrated value of the acceleration component α of the vehicle 1 is increased. Therefore, when the angle measurement error is large, a sufficiently large maneuver of the vehicle 1 is required, and the observability cannot be uniformly determined.

The information matrix F of the equation (9) is defined for the true value vX ₀ of the internal state quantity, and cannot be used in the actual target motion analysis. Therefore, it is an inverse matrix of the estimation error covariance matrix of the state quantity vX ₀ shown in the following equation (10) using the initial value # (vX ₀ ) of the analysis obtained by solving the pseudo linear equation of the equation (7). The estimated information matrix #F is obtained and used instead of the true information matrix F for observability determination.

[Equation 7] This estimation information matrix #F corresponds to a gradient when solving the nonlinear equation of equation (6) using the initial value # (vX ₀ ), and further, the estimation information matrix #F of equation (11) below is used. Whether the nonlinear equation of equation (6) can be solved by obtaining the ratio γ between the maximum eigenvalue and the minimum eigenvalue and comparing it with a predetermined threshold, that is, whether observability is satisfied Judgment is possible. γ = maximum eigenvalue / minimum eigenvalue (11) In the above description, as an example, the internal state quantity at the time of analysis is XY
Although expressed using the coordinate system, other coordinate systems such as X-
The same applies when a YZ coordinate system, polar coordinates, modified polar coordinates, and the like are used. In addition to the usual eigenvalue calculation method, the information matrix is a symmetric positive definite matrix,
An eigenvalue calculation method using singular value decomposition that utilizes the fact that the eigenvalue and the singular value match may be used. Further, a method of obtaining only the maximum eigenvalue and the minimum eigenvalue without obtaining all the eigenvalues, or estimating the distribution range of the eigenvalues without calculating the eigenvalues, and calculating the ratio γ between the maximum eigenvalue and the minimum eigenvalue of the equation (11) By the method of calculating, the eigenvalue is calculated efficiently with reduced calculation amount.

The method for calculating the above eigenvalue is described in the following document. Reference: GHGolub & C.F.Van Loan, “Matrix Computation
s ”2nd ed. The Jhons Hopkins University Press, 199
1 FIG. 1 is a functional block diagram of a desired motion analysis apparatus for carrying out a desired motion analysis method according to an embodiment of the present invention. Elements common to those in the conventional FIG. 3 are designated by common reference numerals. ing. This target motion analysis device has an input terminal IN1 for inputting a signal in from a wave receiver sensor array (not shown) attached to the vehicle 1 in FIG. The input terminal IN1 receives the signal in and receives the measurement result of the azimuth angle θ (t) of the target 2 in FIG. 2, which is the azimuth observation amount β (t) and the error standard deviation σ (of the azimuth observation amount β (t). It is connected to the input side of the azimuth information calculation unit 12 that calculates t) in time series.
The output side of the azimuth information calculation unit 12 is connected to the input side of an initial value calculation unit 14A that calculates an initial value # (vX ₀ ) for motion analysis from the azimuth observation amount β (t). Further, this target motion analysis device is configured such that the acceleration signal α of the navigation vehicle 1 input from the acceleration signal input terminal IN2, the initial value # (vX ₀ ), the azimuth observation amount β (t), and the azimuth observation amount β (t). Error standard deviation σ
An information matrix calculator 15 is provided which inputs (t) and calculates an estimated information matrix #F which is an inverse matrix of the estimated error covariance matrix of the internal state quantity vX ₀ of the target 2. Information matrix calculation unit 15
The output side of is connected to the input side of the eigenvalue calculation unit 16 that calculates the eigenvalue of the estimated information matrix #F. The output side of the eigenvalue calculation unit 16 is an observability determination signal S which is a control signal for starting the motion analysis of the target 2 based on the eigenvalue.
It is connected to the input side of the observability determination unit 17 that generates 17. The output side of the observability determination unit 17 has an initial value # (v
X ₀ ), the azimuth observation amount β (t), and the error standard deviation σ (t) of the azimuth observation amount β (t) are input, and the observability determination signal S1 is input.
Estimated value # {vx ₂ (t)} of the position vector vx ₂ (t) of the target 2 which is the internal state quantity of the target 2 and estimated value # of the velocity vector vx ^* ₂ (t) of the target 2 based on 7. {Vx
It is connected to the internal state quantity estimation unit 18A that calculates ^* ₂ (t)}. On the output side of the internal state quantity estimating unit 18A, the estimated values # {vx ₂ (t)}, # {vx ^* ₂ (t)} are targeted 2.
It is connected to the output terminal OUT which outputs as a result of the motion analysis of.

Next, a target motion analysis method of this embodiment using the target motion analysis device of FIG. 1 will be described. The signal in from the signal source (not shown) received by the receiver sensor array is
Input to the azimuth information calculation unit 12 via the input terminal IN1.
The azimuth information calculation unit 12 receives the signal in and receives the azimuth observation amount β (t) and the error standard deviation σ of the azimuth observation amount β (t).
Calculate (t). The initial value calculation unit 14A uses t = t ₁ ,
..., n sets of azimuth observation amount β (t) at t _n are input, and the initial value of motion analysis # (vX ₀ ) is calculated based on the equation (7).
And outputs the result to the information matrix calculation unit 15 and the internal state quantity estimation unit 18A. The information matrix calculator 15 inputs the azimuth observation amount β (t), the error standard deviation σ (t) of the azimuth observation amount β (t), the acceleration signal α, and the initial value # (vX ₀ ), and formula (10) is used. To calculate the estimated information matrix #F.
The eigenvalue calculator 16 inputs the estimated information matrix #F and calculates the eigenvalue of the estimated information matrix #F. Observability determination unit 17
Is the ratio γ between the maximum eigenvalue and the minimum eigenvalue shown in equation (11).
Is calculated and compared with a predetermined threshold value. If this ratio γ is smaller than the threshold value, it is observable. Otherwise, it is determined as non-observable. If observable, it is possible to start the motion analysis. The observability determination signal S17 is sent to the internal state quantity estimation unit 18A. The internal state quantity estimation unit 18A uses the observability determination signal S1.
When 7 is input, the nonlinear equation shown in the equation (6) is solved using the initial value # (vX ₀ ) as the initial value of the motion analysis, and the target 2
Position vector vx ₂ (t) of target 2 which is the internal state quantity of
Estimate _{# {vx 2 (t)}} , and obtains an estimate of the velocity vector of the target _{^{2 vx * 2 (t) #}} {vx * 2 (t)}, from the output terminal OUT.

As described above, in this embodiment, the azimuth observation amount β (t), which is the measurement result of the azimuth angle θ (t) of the target 2, and the error standard deviation σ (t of the azimuth observation amount β (t). ), And the integrated value of the acceleration signal α of the navigation vehicle 1,
By solving the quasi-linear equation (7) obtained as follows, the estimated information matrix #F is calculated using the obtained initial value # (vX ₀ ) of the motion analysis, and the estimated information matrix #F is calculated from the eigenvalues of the estimated information matrix #F. Since the start point of the motion analysis of the target 2 is controlled by determining the observability, compared with the conventional method of controlling the start time of the motion analysis of the target 2 by using the maneuver of the vehicle. The timing to start the analysis can be appropriately determined.

[0015]

As described in detail above, according to the first invention, the azimuth observation amount, which is the measurement result of the azimuth angle of the target,
The estimated information matrix is calculated using the initial values of the obtained motion analysis by solving the pseudo-linear equation obtained by setting the observation noise component to 0 from the accuracy of this direction observation amount and the integrated value of the acceleration component of the navigation vehicle. However, since the observability is determined from the eigenvalues of this estimated information matrix to control the start time of the target motion analysis, the start time of the target motion analysis is determined using the conventional maneuver of the vehicle. Compared to the control method, the timing for starting the motion analysis can be appropriately determined.
According to the second invention, the eigenvalue of the first invention is calculated by using the eigenvalue calculation method using the singular value decomposition, so that the motion analysis of the target is started using the conventional maneuver of the vehicle. Compared with the method of controlling the time point, the timing for starting the motion analysis can be appropriately determined. According to the third invention, when the eigenvalue of the first invention is calculated, only the maximum eigenvalue and the minimum eigenvalue are obtained, so that it is possible to calculate the eigenvalue with high efficiency while reducing the calculation amount. According to the fourth invention, the eigenvalue of the first invention is not calculated, but the distribution range of the eigenvalues is estimated to obtain the ratio between the maximum eigenvalue and the minimum eigenvalue. A good eigenvalue can be calculated.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a target motion analysis device for carrying out a target motion analysis method according to an embodiment of the present invention.

FIG. 2 is a geometrical explanatory diagram showing an observation system and a motion system in the target motion analysis method.

FIG. 3 is a functional block diagram of a desired motion analysis device for carrying out a conventional desired motion analysis method.

[Explanation of symbols]

 12 azimuth information calculation unit 14A initial value calculation unit 15 information matrix calculation unit 16 eigenvalue calculation unit 17 observability determination unit 18A internal state quantity estimation unit

Claims (4)

[Claims] 1. An azimuth observation amount which is a measurement result of the azimuth angle when a sound emitted from the target is received by a receiver sensor array attached to a movable vehicle and the azimuth angle of the target is measured. And azimuth information calculation processing for calculating the accuracy of the azimuth observation amount in time series, and observation of the azimuth observation amount from the azimuth observation amount, the accuracy of the azimuth observation amount, and the integrated value of the acceleration of the navigation vehicle. An initial value calculation process of calculating an initial state quantity for performing a motion analysis of the target by solving a quasi-linear equation obtained with a noise component as 0, and the initial state quantity as an initial value of the target motion analysis, An internal state quantity estimation process of obtaining an estimated value of the state quantity of the target based on a time series of the observed quantity and a time series of the accuracy of the azimuth observed quantity, in the target motion analysis method for sequentially applying the azimuth observed quantity, Accuracy of azimuth observation amount, An information matrix calculation process of calculating an estimated information matrix that is an inverse matrix of the estimation error covariance matrix of the target state quantity based on the integral value of velocity and the initial state quantity, and an eigenvalue that calculates an eigenvalue of the estimated information matrix. A calculation process and an observability determination process for generating an observability determination signal that is a control signal for starting the motion analysis of the target based on the eigenvalue are performed, and the observability determination signal is generated based on the observability determination signal. A target motion analysis method characterized in that an internal state quantity estimation process is performed. 2. The target motion analysis method according to claim 1, wherein whether or not to start motion analysis of the target is determined based on an eigenvalue obtained by singular value decomposition of the information matrix. 3. The target motion analysis method according to claim 1, wherein only the maximum eigenvalue and the minimum eigenvalue are obtained to determine whether to start the motion analysis of the target. 4. The target motion analysis method according to claim 1, wherein the distribution range of the eigenvalues is estimated, and whether or not to start the motion analysis of the target is determined based on the estimated distribution range. .