KR101097169B1 - System and method for estimating location of mobile object - Google Patents

Abstract

The present invention is a position prediction system of a moving object for predicting a future position of the moving object from the moving object information calculated by sensing the movement of the moving object, the tracking filter having a plurality of filters each including a plurality of dynamic models, and the filters And a predictive modeling unit for predicting the future position of the moving body by combining the probabilities detected from the probability control unit by a probability control function, and the probabilities output from the probability control unit, wherein the probability control function comprises a signal. Disclosed are a sigmoid function or a modified sigmoid function in which a linear function and a sigmoid function are combined.

Description

Position Prediction System and Position Prediction Method of Moving Object {SYSTEM AND METHOD FOR ESTIMATING LOCATION OF MOBILE OBJECT}

The present invention relates to a position prediction system and a position prediction method of a moving object for predicting a future position of the moving object.

In general, guns mounted on military vessels are for firing shells and hitting targets. If the target is stationary, aim the gun at the gun and fire the shell, but if the target moves, you must predict the future position of the target and fire the shell.

In military ships, a fire control system is applied to predict the future position of a moving target and provide this information to the gun. Such a fire control system includes a tracking sensor, a fire information calculator, and a gun.

The target information sensed by the tracking sensor is input to the shooting information calculating device, and the shooting information calculating device performs the calculation to predict the future position of the target and applies this information to the gun.

In recent years, various attempts have been made to improve the accuracy of prediction in the prediction of the target's future position by the shooting information calculator.

The present invention is to provide a position prediction system and a position prediction method of a moving object in which the prediction accuracy is improved in predicting the future position of the moving object.

The present invention for realizing the above object is a position estimation system of a moving object for predicting the future position of the moving object from the moving object information calculated by sensing the movement of the moving object, having a plurality of filters each containing a plurality of dynamic models A predictive modeling unit for predicting a future position of the moving object by combining a tracking filter, a probability control unit for controlling the probabilities detected from the filters by a probability control function, and a probability output from the probability control unit. The probability control function is a sigmoid function or a modified sigmoid function combining a linear function and a sigmoid function.

On the other hand, the present invention detects moving object information by sensing the movement of the moving object, and calculates the probabilities for the dynamic characteristics of the target through a plurality of filters each comprising a plurality of dynamic models based on the moving object information And controlling the calculated probabilities through a probability control function, and combining the controlled probabilities to predict a future position of the moving object, wherein the probability control function comprises a sigmoid function. Or a modified sigmoid function in which a linear function and a sigmoid function are combined.

The present invention having the above configuration can improve the accuracy in predicting the future position of the moving object using the sigmoid function or the modified sigmoid function as the probability control function, and can maintain the persistence and stability of the tracking performance. .

EMBODIMENT OF THE INVENTION Hereinafter, the position prediction system of a mobile body and the position prediction method of a mobile body which concern on this invention are demonstrated in detail with reference to drawings.

Hereinafter, the fire control system will be described as an example, and the fire information calculating device and the fire information calculating method applied to the fire control system can be applied to both the position prediction system and the position prediction method for predicting the future position of the moving object. will be.

1 is a block diagram of a fire control system according to an embodiment of the present invention.

Referring to FIG. 1, the fire control system includes a tracking sensor 100, a shooting information calculating device 200, and a gun 300.

The tracking sensor 100 senses the movement of the moving object (target) to calculate the moving object (target) information (eg, the position, direction, speed, acceleration, etc. of the moving object).

The shooting information calculating device 200 calculates shooting information for controlling the gun 300, that is, gun command, based on the target information from the tracking sensor 100, and the gun 300 calculates the calculated turning angle and elevation. It will operate in response to gun command.

The shooting information calculator 200 may include a prefilter unit 210, a tracking filter 220, a predictive modeling unit 230, a ballistic equation calculator 240, a coordinate converter 250, and an interface. It may include a portion 260.

When the target information acquired by the tracking sensor 100 enters the shooting information calculating apparatus 200, the noise is removed by the prefilter unit 210, and the tracking filter 220 estimates the starting state of the target.

The prediction modeling unit 230 predicts a future position of the target based on the information calculated by the tracking filter 220. The ballistic equation calculator 140 calculates ballistic equations based on the predicted future position of the target and calculates shooting information (ie, turning angle of the gun, elevation, etc.) for firing the shell to the corresponding position. Such shooting information may be calculated based on North Horizontal Coordinates.

The coordinate converting unit 250 converts the true north reference coordinate system reference value into a hull reference coordinate system reference value based on attitude information of the ship on which the gun is mounted. The converted value is input to the interface unit 260, and the interface unit 260 matches an interface for input / output for each gun having various types.

The gun command matched by the interface unit 260 is applied to the gun 300, and the gun 300 fires a shell to a corresponding position by the gun gun command.

2 is a block diagram illustrating a detailed configuration of a tracking filter and related configurations.

The tracking filter 220 is for estimating the activation state of the target, and is implemented in the form of an assembly including a plurality of filters 221a, 221b, ..., 221n.

The kinetic models included in each of the filters 221a, 221b,..., 221n included in the tracking filter 220 include dynamic characteristics such as constant velocity, equivalent acceleration, and circular motion of the target. The tracking filter 220 performs tracking by assuming that the maneuvering of the target to be tracked has the dynamic characteristics of each dynamic model.

The tracking filter 220 may include only a single filter having a single model, but in this case, the tracking filter 220 has an ideal result when the target maneuvering matches the filter's dynamic model, but otherwise the performance of the filter is degraded. Have In the present invention, the tracking filter 220 is configured as a plurality of filters 221a, 221b, ..., 221n having different kinetic models. This allows filtering of various dynamic characteristics of the target, resulting in an excellent overall estimation.

Prediction modeling by the predictive modeling unit 230 is required to predict a future position of the number target filtered by the tracking filter 220. In predictive modeling, the same kinetic characteristics (equivalent velocity, equivalent acceleration, circular motion, etc.) used in the tracking filter 220 are used.

When performing predictive modeling, one of a plurality of dynamics models is selected or a plurality of dynamics models are combined according to a probability obtained based on a covariance result of each filter 221a, 221b, ..., 221n.

In the case of selecting and using only the most probabilistic dynamics models, the effect of using a single filter is obtained. When the target maneuvers coincide with the dynamics model of the predictive modeling unit 230, the result is good, but the probability between the dynamic models If they intersect or the probability difference is small, there is a problem that the performance is degraded.

In the case of using a linear combination of multiple kinetic models according to the probability, the performance where the probability intersects is better than using a single kinetic model, but the filters 221a, 221b, ..., 221n are stable. Even after convergence, if the probability is not high enough, the actual maneuver and other maneuver models may affect the results and reduce the predictive performance.

In the present invention, when the result values of the plurality of filters 221a, 221b, ..., 221n are used in combination, the probability provided by the filters 121a, 121b, ..., 121n is changed without using them. By using in combination, the prediction performance was improved. To this end, a probability control unit 240 is provided between the tracking filter 220 and the predictive modeling unit 230.

The probability control unit 240 controls the probabilities detected from the filters 221a, 221b, ..., 221n by the probability control function. A method of controlling the probabilities by the probability control function will be described later.

A probability normalization unit 150 may be additionally provided between the probability control unit 240 and the prediction modeling unit 230, and the probability normalization unit 250 normalizes the probabilities of the models for which the control is completed. In other words, the probability normalization unit 250 modifies the sum of the probabilities to maintain 100% so that normalized probabilities can be used in predictive modeling.

Finally, the predictive modeling unit 260 combines the controlled probabilities to predict the future position of the target.

3 is a flowchart illustrating a method of calculating fire information associated with an embodiment of the present invention.

First, the tracking sensor 100 detects the target information by sensing the movement of the target (S10), and the tracking filter 220 calculates positional information and probabilities for each dynamic model based on the target information (S20).

In addition, the probability control unit 240 controls the model-specific probabilities calculated by the tracking filter 220 through the probability control function (S30). The probability normalization unit 250 normalizes the probabilities of the models for which the control is completed (S40), and the prediction modeling unit S50 predicts a future position of the target by combining the normalized probabilities, and then calculates the ballistic equation calculation unit 240. Is applied.

Hereinafter, a method of controlling the probability of each model calculated by the tracking filter 220 through the probability control function will be described.

When the measured value from the tracking sensor 100 is filtered through the tracking filter 220, the number of results of the filters 221a, 221b,..., 221n is calculated, and the outputs are combined to perform predictive modeling. .

4 to 6 are graphs showing an example of results obtained by combining the probabilities produced by the filters.

These graphs illustrate the case where there are two models (Model 1, Model 2). In these graphs, the horizontal axis represents input values for probability control, and the vertical axis represents output values for probability control.

4 shows an example of a linear combination using a linear function as a probability control function. As described above, in the case of linear combination, the performance where the probability crosses is better than using a single kinetic model, but the probability remains even after the filters 221a, 221b, ..., 221n converge stably. If it is not large enough, actual maneuvers and other maneuver models may affect the results and degrade predictive performance.

In order to solve this problem, the sigmoid function is used as the probability control function. Figure 5 shows the results of using the sigmoid function as a probability control method according to an embodiment of the present invention.

If the probability is controlled using the sigmoid function, it is made smaller when the probability is small and is made larger when the probability is large. This has the effect of making the probability of a maneuver model with a large probability of actual maneuver bigger, and making the probability of other models smaller, so that it has the predicted value of the model close to the real maneuver.

As can be seen in the graph of Figure 5, the sigmoid function takes the form of 'S' has the effect of increasing the probability difference between the models. Therefore, the probability control method using the sigmoid function has better performance than the linear combination.

6 shows a result of using a modified sigmoid function that complements the sigmoid function as a probability control method according to another embodiment of the present invention.

When using the sigmoid function, since the probability is controlled over the entire interval, the problem does not occur in the part where the probability difference is large. However, it is possible to amplify the error in the part where the probability difference is not large because one model does not converge correctly.

The sigmoid function increases the probability that a model with even a small probability is an accurate model, and therefore, the probability control in the non-converged state may determine that the inaccurate model is accurate and may make the result of predictive modeling unstable. There is a possibility.

In order to minimize this possibility, in this embodiment, a modified sigmoid function combining a linear function and a sigmoid function is used.

In the modified sigmoid function, the linear function is maintained in the section determined not to converge to one model, and the sigmoid function is used to amplify the difference of probability significantly in the other sections. As shown in FIG. 6, a section to which a linear function is applied is referred to as a 'linear holding section' and is referred to as a 'sigmoid section' to which a sigmoid function is applied to both sides of the 'linear holding section'.

Here, a differential continuous value was created between the linear maintenance interval and the sigmoid interval to maintain the stability of the prediction result when predictive modeling.

Using the modified sigmoid function, when the results from the plurality of filters 221a, 221b, ..., 221n do not converge, the linearity is maintained and the probability difference between the models is increased in other sections. . Accordingly, it is possible to maintain the persistence and stability while improving the accuracy of the prediction result in the prediction modeling.

As such, the present invention improves accuracy while maintaining the persistence and stability of tracking performance through the probability control method as described above.

The position prediction system and the position prediction method of the moving object described above are not limited to the configuration and method of the embodiments described above, but the embodiments may be selectively combined with all or some of the embodiments so that various modifications may be made. It may be configured.

1 is a block diagram of a fire control system according to an embodiment of the present invention.

2 is a block diagram illustrating a detailed configuration of a tracking filter and related configurations.

3 is a flowchart illustrating a method of calculating fire information associated with an embodiment of the present invention.

4 is a graph showing the result of probability control using a linear combination.

Figure 5 is a graph showing the results of using the sigmoid function as a probability control method according to an embodiment of the present invention.

6 is a graph showing a result of using a modified sigmoid function as a probability control method according to another embodiment of the present invention.

Claims (9)

delete In the position prediction system of the moving object for predicting the future position of the moving object from the moving object information calculated by sensing the movement of the moving object, A tracking filter having a plurality of filters each including a plurality of dynamic models; A probability control unit for controlling the probabilities detected from the filters by a probability control function; And A prediction modeling unit for predicting a future position of the moving object by combining the output probabilities from the probability control unit, The probability control function is a modified sigmoid function in which a linear function and a sigmoid function are combined, and the probabilities detected from the plurality of filters of the tracking filter are respectively a dynamic model of one of the plurality of dynamic models. The mobile body is characterized by using a linear function when it is in a section that is difficult to determine, and using a sigmoid function when it is in a section that can be determined by one of the plurality of dynamic models. Position prediction system. The method of claim 2, wherein the resultant value of the modified sigmoid function is A linear maintenance interval to which the linear function is applied; And And a sigmoid section to which the sigmoid function is applied to both sides of the linear maintenance section. The method of claim 3, And a differential succession between the linear maintenance section and the sigmoid section. delete Sensing moving object information to detect moving object information; Calculating probabilities for the dynamic characteristics of the moving object through a plurality of filters each including a plurality of dynamic models based on the moving object information; Controlling the calculated probabilities through a probability control function; And Combining the controlled probabilities to predict a future position of the moving object, The probability control function is a modified sigmoid function in which a linear function and a sigmoid function are combined, and the probabilities detected from the plurality of filters are determined as a dynamic model of one of the plurality of dynamic models. In a difficult section, a linear function is used, and in a section in which a dynamic model of one of the plurality of dynamic models can be determined, a sigmoid function is used to control the probability of the moving object. Way. The method of claim 6, wherein the resultant value through the modified sigmoid function is A linear maintenance interval to which the linear function is applied; And And a sigmoid section formed on both sides of the linear maintenance section, to which the sigmoid function is applied. The method of claim 7, wherein And a differential succession between the linear maintenance section and the sigmoid section. The method of claim 6, And a normalization step of normalizing output values of the probability control function.

Images