JPH0690278B2 - Tracking filter for radar - Google Patents

Description

The present invention relates to a radar tracking filter, and more particularly to a radar tracking filter suitable for tracking a target by an electronic scanning antenna.

[Conventional technology]

When tracking a target with an electronic scanning antenna, the cross-correlation between the radar target information from the previous scan and the target machine information from the next scan is examined for each target machine, and from among many target machine information inputs, The process of determining the time transition state of the target aircraft is performed. By this processing (scan correlation processing), the position of each target aircraft for each scan is associated one-to-one as a track, and identification is performed. And by continuing this identification,
The target is tracked.

By the way, the position information of the target aircraft obtained by the radar contains various measurement errors. To reduce these measurement errors and improve tracking performance, the target aircraft's position and motion conditions are adjusted according to the motion state and measurement state of the target aircraft. Smoothing is required.

In simultaneous search / tracking of multiple target aircraft with a search radar used for air traffic control, if it is known that there is only one target aircraft within the radar coverage area, the identification and tracking of the target is obviously extremely difficult. It's easy. That is, identifying the goal,
The essence of tracking is how to identify and maintain identification of multiple aircraft in close proximity. Specifically, how to make a prediction gate of the position at the time of the next scan for each tracking target aircraft and how to reduce the size of the gate to achieve both narrow tracking performance and separation performance from other targets well Can be seen as a problem. In the calculation, the aim is to capture the true track quickly and accurately by using the radar observation position sample data for each fixed period (antenna scan), and the following basic operations are required.

． The sample data obtained up to now are accumulated for each goal.

． Smooth the dispersion of data and predict the sample data position for the next scan.

． Set a tracking gate that considers the allowable range of the difference between the predicted position data and the actually input observed position data, and avoiding the risk of multiple correlation.

． A predetermined correlation detection calculation is executed for multiple correlations.

Next, tracking using an α-β filter, which is widely used for tracking a target machine by an electronic computer, will be described. FIG. 5 shows a conceptual diagram of tracking using this α-β filter. In FIG. 5, the position smoothing coefficient is α and the velocity smoothing coefficient is β.
As a straight line predictor, the nth scan to the (n +
1) It shows how to predict a scan.

That is, assuming that the predicted position in n scans is P (n) and the observed data position is R (n), the tracking error E (n) is E (n) = R (n) -P (n) ... ……… (1)
Is.

The smooth position S (n) and the smooth velocity (n) are as follows: S (n) = P (n) + αE (n) (2)
(N) = (n-1) + βE (n) / T (3)
Required by. Here, T is a sampling period. From now on, P (n +) indicating the predicted position of (n + 1) scan
1) is as follows: P (n + 1) = S (n) + (n) · T (4)
Required by.

α = 0 indicates a case where the predicted position is a smooth position, and α = 1
Indicates the case where the observation data position is a smooth position. It means that the smaller α is, the deeper the smoothing is. Therefore, when it is determined that the motion is a uniform linear motion, both α and β are reduced to sufficiently smooth the fluctuation, thereby improving the tracking accuracy. The values of the smoothing parameters α and β take into consideration the stability of the solution of the recurrence formula and the transient response characteristics (including the followability to the turning of the target aircraft), as well as the effects of missing data due to radar detection omissions. Need to be set.

Here, assuming the target model formula, α-
The β filter will be described in more detail.

The target model formula converted from (R, θ) coordinates to (X, Y) coordinates in the radar tracking process is assumed as follows.

(I). Coordinate conversion formula “X (n) = R (n) sin θ (n) Y (n) = R (n) cos θ (n)” ……………… (5)
(Ii). System state equation (constant velocity model) “X (n + 1) = X (n) + TV _x (n) Y (n + 1) = Y (n) + TV _y (n) V _x (n + 1) = V _x (n) + ω _x (n) V _y (n + 1) = V _y (n) + ω _y (n) ”……… (6)
(Iii). Observation equation “X _M (n) = X (n) + υ _x (n) Y _M (n) = Y (n) + υ _y (n)” ……………… (7)
Here, n is the observation time (scan), R (n) is the target range at n time, θ (n) is the azimuth (azimuth) of the target at n time, and X (n) is at n time. Target X coordinate position, Y
(N) is the target Y coordinate position at time n, T is the observation time interval (scan time interval), V _x (n) is the target X at time n
Directional speed, V _Y (n) is the target Y-direction speed at time n, ω
_x (n) is the system disturbance applied in the target X direction at the time n, ω _y (n) is the system disturbance applied in the target Y direction at the time n, X _M (n) is the target X at the time n Coordinate observation position, Y _M (n) is the target Y coordinate observation position at time n, υ _x
(N) indicates an observation disturbance added to the target X-coordinate position observation at time n, and υ _y (n) indicates an observation disturbance added to the target Y-coordinate position observation at time n.

However, E [ω _x (n)] = 0 E [ω _x (i) ω _x (j)] = Q _x δij E [ω _y (n)] = 0 E [ω _y (i) ω _y (j )] = Q _y δij E [υ _x (n)] = 0 E [υ _x (i) υ _x (j)] = R _x δij E [υ _y (n)] = 0 E [υ _y (i) υ _y (j)] = R _y δij where Q _x , Q _y , R _x and R _y are covariances, δij is the Kronecker delta, that is, δij = 1 when i = j, and δij when i ≠ j. = 0 Taking the target model given by the above equations (5) to (7) as an example, from the equation (4), (n + 1 | n) = (n | n) + T · _x (n | n) ... … (8a)
(N + 1 | n) = (n | n) + T · _Y (n | n) …… (8b)
Rather than being a constant velocity motion model, _x (n + 1 | n) = _x (n | n) ……………………… (9a)
_Y (n + 1 | n) = _Y (n | n) ……………………… (9b)
Further, from the formulas (1) and (2), (n | n) = (n | n-1) + α {X _M (n)-(n | n-1)} ...
…… (10a) (n | n) = (n | n-1) + α {Y _M (n)-(n | n-1)} ……
...... (10b) From equations (1) and (3), V _x (n | n) = V _x (n | n-1) + β / T {X _M (n) -X (n | n-1) }…
(11a) V _Y (n | n) = V _Y (n | n-1) + β {Y _M (n) -Y (n | n-1)} / T ...
(11b) where β = α ² / (1-α) (α and β are smoothing coefficients and constants)
(12) Also, (n + 1 | n) is the next time point X (n
+1) predicted value, where (n | n) is X at the time point n
The smoothed estimated value of (n) is shown.

FIG. 4 is a block diagram showing the equations (8a), (9a), (10a), (11a) (X direction). By giving the initial condition (1 | 1) = X _M (1) (1 | 1) = {X _M (1) −X _M (0)} / T, the prediction estimation by the α-β filter is recursive. Sought.

The reason why there are two output signals in FIG. 4 is that it is necessary to extract the predicted value of the position in the X direction and the predicted value of the X component of the velocity that have been smoothed by the smoothing coefficients α and β, respectively.

In FIG. 4, the transfer function α of the position smoothing processing unit 11 and the transfer function β / T of the velocity smoothing processing unit 12 are constants. The delay units 3 and 4 are configured to output a value one observation time before (T time).

Further, in FIG. 4, an X coordinate filter is shown. The Y-coordinate filter has the same structure, and X in FIG. 4 is replaced with Y.

The value of α is determined depending on which of the two contradictory items, that is, noise removal and target followability, is emphasized. In the conventional technique, the value of α is predetermined with an appropriate constant of 0 to 1. β was also a constant that was uniquely determined by “β = α ² / (1-α)”.

If the angle measurement error l and the distance measurement error l _R are set, the error l _x of the X position after conversion into XY coordinates is given by the following equation.

l _x = R ・ | sin (θ ＋ l _θ ) −sin _θ ｜ ＋ l _R・ | sin θ ｜ = R ・ | sin _θ cosl _θ ＋ cos θsinl _θ −sin _θ ｜ ＋ l _R・ | sin
θ | = R ・ | sin θ (cosl _θ −1) + cos _θ sinl _θ ｜ ＋ l _R・ | si
nθ | ·········· (12a) Here, for example, if “l _R = 0.02 degree”, cosl = 0.999999939≈1 can be set.

Therefore, l _x = R ・ | cos _θ sinl _θ ｜ ＋ l _R・ | sin θ ｜ = 0.00035 ・ R ・ | cos θ ｜ ＋0.005 ・ | sin θ ｜ ……………
(12a ′) Therefore, the magnitude of the observed disturbance converted into the X coordinate is a function of R and θ. That is, the magnitude of the disturbance changes depending on the target position.

For example, if l _R = 0.05 NM (Nautical Mile: Nautical Miles), then l _x = 0.0875 NM when R = 250 NM and θ = 0 °, and l _x = 0.005 when R = 250 NM and θ = 90 ° NM, further, when R = 5 NM and θ = 0 °, l _x = 0.00175 NM and the value of l _x greatly changes with the value of R and θ.

[Problems to be solved by the invention]

First, in order to clarify the problem, it will be described with reference to FIGS. 2 and 3 whether the value of α is 0 or 1 for which the observation disturbance is large and the observation disturbance is small.

As shown in FIG. 3, when the observed disturbance is large, if the target is moving at a substantially constant velocity, the smaller α is, the more noise can be removed, and a better prediction can be made. However, considering the followability to the turning motion to some extent, α is actually at least 0.5
Compromising on the degree can give a good overall prediction. On the contrary, when the observed disturbance in Fig. 3 is small, α
The larger is, the better the followability and the better the estimation. Since the disturbance is originally small even when the vehicle is not turning, a good estimate can be made even if α is large.
It is understood that the larger is, the better.

However, in the above-mentioned conventional example, the values of the position smoothing coefficient α and the velocity smoothing coefficient β are predetermined as appropriate constants, even though the magnitude of the observed disturbance greatly changes depending on the target position. There is a portion where the prediction accuracy is extremely deteriorated depending on the position of the target, which causes a problem that the tracking cannot be performed comprehensively with high accuracy.

[Object of the Invention]

The present invention improves the inconveniences of the conventional example, reduces the variation in the extreme prediction accuracy due to the target position, and provides a radar tracking filter that enables highly accurate tracking overall. To that end.

[Means for solving problems]

In the present invention, a position smoothing processing circuit that performs a position smoothing process for a tracking error of radar observation, and a speed smoothing processing circuit that performs a speed smoothing process for a tracking error of radar observation are provided, and based on the output of each of these circuits. In a tracking filter for radar that performs a predetermined calculation and outputs a predicted position value and a predicted speed value of a target in orthogonal coordinates, a smoothing processing circuit and a speed smoothing processing circuit perform smoothing processing on radar observation information in a position smoothing processing circuit and a speed smoothing processing circuit. A smoothing coefficient calculation circuit for calculating and outputting a position smoothing coefficient or a velocity smoothing coefficient required for is additionally provided.

This smoothing coefficient calculation circuit inputs the absolute value of the distance data and azimuth data obtained by radar observation and calculates the observation error on the position coordinates, and this observation error calculation function operates. When the observation error is calculated and the maximum value is calculated, the position smoothing coefficient maximum value setting function for setting the position smoothing coefficient corresponding to the maximum value to a predetermined value around 0.5 is provided. At the same time, the smoothing coefficient calculating circuit has a function of calculating the above-described position smoothing coefficient and velocity smoothing coefficient within the range set by the above-described position smoothing coefficient maximum value setting function. This aims to achieve the above-mentioned object.

[Action]

The value of the smoothing coefficient α when the maximum error of the absolute value of the observed value data is calculated is set to 0.5, which improves the follow-up during turning at the maximum error and at the same time smooths at the maximum error. Since the coefficient α of 0.5 is used to discriminate all other situations, the observation disturbance is largely removed at the position where the observation disturbance is large, and overall accurate prediction can be made and the observation disturbance is small. At the position, particularly when the target is turned, the followability with respect to the true value of the predicted value is improved, and it is possible to perform accurate prediction overall.

Example of Invention

An embodiment of the present invention will be described below with reference to FIG.
Here, the same reference numerals are used for the same parts as those in FIG. 4 in the above-mentioned conventional example.

In FIG. 1, transfer functions α _X (n) and β _X (n) / T of the position smoothing processing unit 1 and the velocity smoothing processing unit 2 are determined by the smoothing coefficient calculation circuit 5.

That is, in the smoothing coefficient calculation circuit 5, first, based on the observed distance data R _M (n) and azimuth angle data θ _M (n), the observation error l _X (n), l for the X coordinate and the Y coordinate is calculated. _{Find Y} (n). In this case, the observation errors l _X (n) and l _Y (n) are determined by the absolute values of the distance data R _M (n) and the azimuth data θ _M (n).

Equation (12a) from _{l X (n) = R M} (n) · | cosθ M (n) · sinl θ | + l R · | sinθ
_{M (n) | ............ (13a} ) l Y (n) = R M (n) | sinθ M (n) · sinl θ | + l R | cosθ M (n)
｜ ………… (13b) Here, taking the monopulse SSR described above as an example, the value determined by the accuracy of the radar is replaced with a number. That is, l = 0.
002 degrees, a l _R = 0.005 nm, equation (13a), are substituted into _{(13b), l X (n} ) = 0.0035 · R M (n) · | cosθ M (n) | +0.005 | sinθ M
(N) | ... (13a ') l _Y (n) = 0.0035R _M (n) | sin θ _M (n) | +0.005 | cos θ
_M (n) | ... (13b ') Then, based on this error, α _X (n) and α _Y (n) are obtained by the following equation, for example.

α _X (n) = C / [C + l _X ² (n)] …………… (14a)
α _Y (n) = C / [C + l _Y ² (n)) …………… (14b)
(C is a constant) The method for determining C is as follows.

First, find the maximum value of the observation error l _x max (R = 250NM
(Maximum radar range), when θ = 0 °).

l _X max = 0.00035 × 250 = 0.0875 (NM) And, if we take the policy that the value of α at the time of maximum error is, for example, α = 0.5, we can calculate from equation (14a) that C = l _X ² max = 0.00766. It suffices (here, α = 0.5 is set in consideration of follow-up during turning at the maximum error).

As a result, α _X (n) and α _Y (n) are respectively calculated by the formula (14
It can be calculated by a) and (14b). And β _X
(N) and β _Y (n) are uniquely determined if α _X (n) and α _Y (n) are obtained, and β _X (n) = α _X ² (n) / {1-α _X (N)} …… (15
a) β _Y (n) = α _Y ² (n) / {1-α _Y (n)} …… (15
b)

As described above, in the smoothing coefficient calculation circuit 5, the equation (13
Calculations are performed based on a '), (14a), and (15a), and the smoothing coefficients α _X (n) and β _X (n) in the X direction are determined by this.

Other configurations are the same as those of the conventional example described above.

Here, although only the X-coordinate filter is shown in FIG. 1, the Y-coordinate filter is also configured in the same manner, and by replacing the X portion in FIG. Is established. The smoothing coefficient calculating circuit of the Y-coordinate filter (not shown) allows the equations (13b ') and (14
By performing the calculation based on b) and (15b), the Y-direction smoothing coefficients α _Y (n) and β _Y (n) can be obtained.

As described above, according to the present embodiment, in the α-β filter in the target model converted from (R, θ) coordinates to (X, Y) coordinates in the radar tracking process, the values of the smoothing coefficients α and β are , By having a calculation circuit that changes and determines by the absolute values of the observed distance data R _M and azimuth data θ _M , the observation disturbance is largely removed at the position where the observation disturbance is large, and overall accuracy is high. There is an advantage that the prediction can be performed, and at a position where the observed disturbance is small, the followability to the true value of the predicted value is enhanced especially at the time of turning of the target, and the overall accurate prediction can be performed.

Here, the distance data R _M and the azimuth angle data θ _M do not necessarily have to be data obtained from radar observation. Also,
In the above embodiment, the smoothing coefficients α and β are not obtained from the variations between the distance data R _M and the azimuth angle data θ _M , but the smoothing coefficients α and β are calculated from the absolute values of the distance data R _M and the azimuth angle data θ _M. It has a feature in the point of seeking.

In the above embodiment, the α-β filter for the constant velocity linear motion model has been described. However, based on the uniform acceleration motion model, the α-β filter including the smoothing coefficient γ of acceleration in consideration of acceleration is also included. Α-β-γ as an extension of
The present invention is also applicable to filters.

〔The invention's effect〕

Since the present invention is configured and functions as described above, according to this, a smoothing coefficient calculation circuit for changing the position smoothing coefficient and the velocity smoothing coefficient in accordance with the absolute values of the separately input distance data and azimuth data is provided. Therefore, it is possible to set the smoothing coefficient to the optimum state in response to the observed disturbance whose values greatly differ depending on the distance and the azimuth angle. Corresponds to the observation error calculation function that inputs the absolute value and calculates the observation error on that position coordinate, and the maximum value when this observation error calculation function operates and the observation error is calculated and the maximum value is calculated A position smoothing coefficient maximum value setting function for setting the position smoothing coefficient to a predetermined value of around 0.5, and a position smoothing coefficient for other situation changes within the range set by this position smoothing coefficient maximum value setting function. And since it is configured to calculate the velocity smoothing coefficient, unlike the above-mentioned conventional example, the maximum value of the position smoothing coefficient α is set to a predetermined value of around 0.5. Disturbances can be significantly removed and highly accurate prediction can be made.Variations are small at positions where observed disturbances are small, and the tracking of the true predicted value at the time of turning of the target is particularly high, resulting in high overall accuracy. Since it is possible to make predictions, it is possible to significantly reduce the variation in the prediction accuracy due to the target position, which has been a problem in the past. Therefore, it is not possible in the past to significantly improve the tracking performance overall. It is possible to provide an excellent tracking filter for radar.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory view of the influence of α when the observation disturbance is large and is performing uniform velocity motion, and FIG. 4 is a block diagram showing a conventional example, and FIG. 5 is a conceptual diagram of tracking using an α-β filter. 1 ... position smoothing processing circuit, 2 ... speed smoothing processing circuit, 5
...... Smoothing coefficient calculation circuit, α, α _X (n) …… Position smoothing coefficient, β, β _X (n) …… Velocity smoothing coefficient, X _M (n) …… X
Direction observation data, R _M (n) …… distance data, θ _M (n)
…… Azimuth data.

Claims (1)

[Claims] 1. A position smoothing processing circuit for smoothing a position with respect to a tracking error of radar observation, and a velocity smoothing processing circuit for performing a velocity smoothing process with respect to a tracking error of radar observation, and based on outputs of these circuits. In a tracking filter for radar that outputs a predicted position value and a predicted speed value on a rectangular coordinate of a target by performing a predetermined calculation, the position smoothing processing circuit and the speed smoothing processing circuit include the smoothing processing circuit for the radar observation. A smoothing coefficient calculating circuit for calculating and outputting a position smoothing coefficient or a velocity smoothing coefficient necessary for smoothing information is provided, and this smoothing coefficient calculating circuit is used for calculating the distance data and the azimuth angle data obtained by the radar observation. The observation error calculation function that inputs the absolute value and calculates the observation error on the position coordinates, and this observation error calculation function operates When the maximum value is calculated, it is equipped with a position smoothing coefficient maximum value setting function that sets the position smoothing coefficient corresponding to the maximum value to a predetermined value around 0.5, and is set by this position smoothing coefficient maximum value setting function. A tracking filter for radar, characterized in that it is configured to calculate the position smoothing coefficient and the velocity smoothing coefficient within a defined range.