Target trajectory estimation device and target trajectory estimation method

Abstract

A target trajectory estimation device is provided with: an orientation information reading unit (41); a reading unit that reads azimuth information indicating azimuth measurement values obtained by each of a plurality of azimuth sensors (10) that receive arrival radio waves from an object target (100), and azimuth measurement error information indicating azimuth measurement errors that are attached to or set in accordance with the azimuth measurement values indicated by the azimuth information; a time domain division unit (43) for dividing the time at which the object target (100) is observed by each of the plurality of orientation sensors (10) into a plurality of time domains in chronological order, and dividing the orientation information read by the orientation information reading unit (41) for each of the divided time domains; and an abnormal value determination unit (44) that determines, for each time domain divided by the time domain division unit (43), whether or not the position indicated by the orientation measurement value indicated by the orientation information present in each time domain is concentrated at one point or around a state vector comprising the position and speed, whether or not the position indicated by the orientation measurement value indicates an abnormal value, and that determines whether or not the position indicated by the orientation measurement value indicates an abnormal value. The azimuth measurement error associated with the azimuth measurement value determined to be an abnormal value is reset to a setting for reducing the reliability of trajectory estimation for the azimuth measurement value, and the azimuth measurement error associated with the azimuth measurement value determined not to be an abnormal value is reset to a setting for reducing the reliability of trajectory estimation for the azimuth measurement value. Resetting so as to maintain the reliability of the trajectory estimation for the azimuth measurement value; a positioning processing unit (45) that estimates, for each time domain divided by the time domain division unit (43), a positioning value in each time domain for the object target (100), using an orientation measurement value indicated by orientation information present in each time domain and an orientation measurement error that has been reset in the orientation measurement value; and an output unit (46) that outputs object trajectory information indicating the trajectory of the object target (100) by connecting the localization values for each time domain estimated by the localization processing unit (45) in time series.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a target trajectory estimation device that estimates a position value of a target and a target trajectory estimation method. BACKGROUND

[0002] In a target trajectory estimation device that estimates a trajectory of a target such as an aircraft based on azimuth measurement values of a plurality of active sensors such as radar sensors or passive sensors such as radio wave sensors, there is a technique that sets an azimuth measurement error based on a reliability of each of the plurality of azimuth measurement values, and thereby calculates a weighted position value.

[0003] For example, there is a method that estimates a position, a velocity, and an accuracy thereof of a target while reflecting a difference in a time of azimuth measurement in a MAP (Maximum A Posteriori) estimation based on different azimuth measurement errors of a plurality of non-synchronous sensors.

[0004] Further, for example, in Patent Literature 1, there is shown a method that performs an outlier determination at a time of a position measurement process of a target in a target trajectory estimation device.

[0005] The target trajectory estimation device shown in Patent Literature 1 calculates a DOP (Dilution Of Precision) that is an index indicating a position accuracy of each of groups of sensors selected from a plurality of sensors, and preferentially sets a sensor used in the position measurement process.

[0006] PRIOR ART DOCUMENTS

[0007] PATENT LITERATURE

[0008] Patent Literature 1: Japanese Patent Application Publication No. 2016-61705 SUMMARY

[0009] PROBLEMS TO BE SOLVED BY THE INVENTION

[0010] The target tracking device shown in Patent Literature 1 has a property that the DOP decreases as the position accuracy deteriorates, in a case where a plurality of sensors exist in a position indicating the same azimuth with respect to a target, and thus, it is possible to perform a good outlier determination at the time of the position measurement process.

[0011] However, when a specific sensor in a sensor group is in a multipath environment, there is a case where an outlier is included in a larger amount, and it is possible that an erroneous position measurement process is performed.

[0012] The present disclosure is made in view of the above points, and aims to obtain a target trajectory estimation device that is capable of estimating a position value of a target in a multipath environment with high accuracy even when a specific sensor is in the multipath environment, and improves trajectory estimation of an object.

[0013] Methods for solving problems

[0014] The target trajectory estimation device disclosed herein includes: an azimuth information reading unit that reads azimuth information representing azimuth measurement values ​​obtained by multiple azimuth sensors that receive arrival radio waves from a target, and azimuth measurement error information representing azimuth measurement errors attached to or uniformly set in the azimuth measurement values ​​shown by the azimuth information; a time domain segmentation unit that divides the time when the multiple azimuth sensors observe the target into multiple time domains in a time-series order, and segments the azimuth information read by the azimuth information reading unit according to each segmented time domain; and an outlier determination unit that, according to each time domain segmented by the time domain segmentation, determines whether the positions represented by the azimuth measurement values ​​shown by the azimuth information in each time domain are concentrated at a single point or concentrated around the periphery of a state vector composed of position and velocity. Whether the azimuth measurement value represents an outlier, the azimuth measurement error attached to the azimuth measurement value that is determined to be an outlier is reset to reduce the reliability of the trajectory estimation for that azimuth measurement value, while the azimuth measurement error attached to the azimuth measurement value that is determined to be not an outlier is reset to maintain the original reliability of the trajectory estimation for that azimuth measurement value; the positioning processing unit estimates the positioning value for the target in each time domain according to each time domain divided by the time domain segmentation part, using the azimuth measurement value shown by the azimuth information present in each time domain and the azimuth measurement error attached to the azimuth measurement value that has been reset; and the output unit outputs object trajectory information that represents the trajectory of the target by connecting the positioning values ​​of each time domain estimated by the positioning processing unit in a time sequence.

[0015] Invention Effects

[0016] According to this disclosure, even when a particular orientation sensor is in a multipath environment, it can accurately estimate the target's position value and perform trajectory estimation that is closer to the target's actual trajectory (true value). Attached Figure Description

[0017] Figure 1 This is a block diagram showing the general structure of a target trajectory estimation system including the target trajectory estimation device of Embodiment 1.

[0018] Figure 2 This is a diagram showing the orientation measurements of multiple orientation sensors segmented in the time domain in the target trajectory estimation device of Embodiment 1.

[0019] Figure 3 It is a diagram showing the orientation measurements of multiple orientation sensors in the target trajectory estimation device of Embodiment 1 in space.

[0020] Figure 4This is a diagram showing the normal and abnormal values ​​within the azimuth measurement values, which are the objects determined by the positioning gate in the target trajectory estimation device of Embodiment 1.

[0021] Figure 5 This is a diagram illustrating the outlier determination method and the gate used in the positioning gate determination in the target trajectory estimation device of Embodiment 1.

[0022] Figure 6 This is a flowchart illustrating an example of the outlier determination process in the target trajectory estimation device of Embodiment 1.

[0023] Figure 7 This is a schematic functional block diagram illustrating an example of the hardware structure of the target trajectory estimation device in Embodiment 1.

[0024] Figure 8 This is a block diagram showing the general structure of the target trajectory estimation device in Embodiment 2.

[0025] Figure 9 This is a flowchart illustrating an example of the process from outlier detection to tracking processing in the target trajectory estimation apparatus of Embodiment 2, which uses a Kalman filter with a positive time direction.

[0026] Figure 10 This is a flowchart illustrating an example of the process from outlier detection to tracking processing in the target trajectory estimation apparatus of Embodiment 2, which uses a Kalman filter with reverse time direction. Detailed Implementation

[0027] Implementation method 1.

[0028] use Figures 1-7 This describes the target trajectory estimation device of Implementation Method 1.

[0029] Figure 1 This is a block diagram showing the general structure of a target trajectory estimation system including the target trajectory estimation device of Embodiment 1.

[0030] like Figure 1 As shown, the target trajectory estimation system has multiple orientation sensors 101 to 101. N Multiple orientation measuring units 201-20 N Storage device 30, target trajectory estimation device 40 and output device 50.

[0031] The target trajectory estimation device 40 of Implementation 1 uses only a plurality of orientation sensors 101 to 102. N The location measurement value, for example, to determine the position without obtaining distance data, and to estimate the trajectory of targets such as aircraft.

[0032] Regarding the subscript N used to mark the orientation sensor 10 and the orientation measuring unit 20, in order to avoid complicated explanations, the subscript will not be used in the following explanations unless a separate explanation is required.

[0033] Multiple (N) orientation sensors 10 are connected and configured to receive signals from targets 100 such as aircraft and obtain their orientation information.

[0034] Each directional sensor 10 is a passive sensor such as an electromagnetic wave sensor that receives electromagnetic waves transmitted from the target 100.

[0035] The multiple orientation sensors 10 have the same structure.

[0036] In addition, each orientation sensor 10 can also be an active sensor such as a radar sensor that outputs signals and receives signals that are hit by the target 100 and reflected.

[0037] In addition, each orientation sensor 10 can also be a sensor capable of sensing and receiving physical quantities such as sound waves or heat emitted from the target 100 and estimating the orientation of the target 100.

[0038] The measurement error of each position sensor 10 can vary depending on each position sensor 10, and can also vary depending on the measurement time.

[0039] Multiple (N) azimuth measuring units 20 respectively measure the azimuth based on the received signals received by the corresponding azimuth sensor 10, and output the measurement results as azimuth information representing the azimuth measurement value.

[0040] Multiple orientation measurement units 20 analyze the corresponding received signals of each of the N interconnected orientation sensors 10 to determine the orientation (θ) of the arriving radio wave from the target 100. k φ k () is used as the orientation value.

[0041] In addition, the multiple orientation measuring units 20 respectively measure the time t of their measurements as information included in the orientation measurement values. k and the position information (x) of the corresponding orientation sensor 10 r k y r k z r k ) and azimuth measurement (θ) k φ k This information is linked together to output as azimuth information representing the measured azimuth value. Azimuth information is digital information.

[0042] Additionally, k is the azimuth measurement value number, used to determine the measurement time t. k Integers from 1 to K (>1).

[0043] The azimuth measurement value number k is a number assigned to all asynchronous azimuth measurement values ​​of target 100 observed by multiple (N) azimuth sensors 10 in chronological order, and is said to have K values.

[0044] In addition, in the location information, x, y, and z represent the x-coordinate, y-coordinate, and z-coordinate, respectively, and r is an integer from 1 to N used to determine the orientation sensor 10.

[0045] Multiple orientation measurement units 20 respectively analyze the orientation measurement error (σ) in a non-multipath environment based on the antenna structure and SNR (signal-to-noise ratio) of the corresponding orientation sensor 10. θ，k σ φ，k This is output as the bearing measurement error information. Bearing measurement error (σ) θ，k σ φ，k ) Accompanying the corresponding azimuth measurement value (θ) k φ k The orientation measurement error information is digital information.

[0046] That is, during the period when the multiple azimuth sensors 10 observe the target 100 respectively, the multiple azimuth sensors 10 receive K arriving radio waves from the target 100 in a time sequence, and the multiple azimuth measurement units 20 obtain K azimuth measurement values ​​(θ) for the target 100. k φ k ) and orientation measurement error (σ θ，k σ φ，k ).

[0047] The storage device 30 stores the orientation information from each of the multiple (N) orientation measuring units 20 in the storage unit 31.

[0048] During the period when the multiple azimuth sensors 10 are observing the target 100, the storage device 30 will send the azimuth measurement values ​​(θ) from the multiple azimuth measuring units 20 in a time sequence. k φ k The azimuth information and the azimuth measurement error (σ) θ，k σ φ，k The azimuth measurement error information is stored in the storage unit 310, and the azimuth information and azimuth measurement error information stored during target trajectory estimation are output.

[0049] The target trajectory estimation device 40 estimates the azimuth value (θ) based on the azimuth information received by multiple azimuth sensors 10. k φ k ), the reset azimuth measurement error (σ) θ，k σ φ，kPosition information of orientation sensor 10 (x) r k y r k z r k ) and representing time t m Estimate the trajectory of targets such as aircraft by 100.

[0050] Specifically, the target trajectory estimation device 40 is a device that can estimate the trajectory of the target 100 with high accuracy even when some of the multiple orientation sensors 10 are in an environment that does not indicate the orientation of the target 100, i.e., a so-called multipath environment.

[0051] The target trajectory estimation device 40 acquires the azimuth measurement values ​​(θ) shown by the K azimuth information from multiple azimuth measurement units 20 stored in the storage device 30. k φ k ), orientation measurement error (σ) θ，k σ φ，k Position information of orientation sensor 10 (x) r k y r k z r k ) and measurement time t k The representative time t is calculated by dividing the time when the target 100 is observed by the multiple orientation sensors 10 into multiple time domains m in a time-series order. m The measured value x m Estimate the representative time t of the time series sequence. m The measured value x m The trajectory of the connected target 100 is output as object trajectory information.

[0052] During the period when multiple orientation sensors 10 observe the target 100, the target trajectory estimation device 40 performs recursive processing to estimate the trajectory based on the orientation measurement values ​​in a certain time domain, thereby obtaining the object trajectory information.

[0053] The output device 50 displays the estimated trajectory of the target 100 obtained from the object trajectory information from the target trajectory estimation device 40.

[0054] The target trajectory estimation device 40 has a azimuth information reading unit 41, an error setting unit 42, a time domain segmentation unit 43, an outlier determination unit 44, a position processing unit 45, and an output unit 46.

[0055] The azimuth information reading unit 41 reads the azimuth measurement values ​​(θ) from multiple azimuth measuring units 20 stored in the storage device 30. kφ k The azimuth information and the azimuth measurement error (σ) stored in the storage device 30 θ，k σ φ，k The orientation measurement error information (σ) θ，k σ φ，k ).

[0056] The azimuth information reading unit 41 reads the azimuth measurement value (θ) stored in the storage device 30. k φ k The location information (x) of the orientation sensor 10 included in the document. r k y r k z r k ) and measurement time t k .

[0057] No indication of orientation measurement error (σ) was obtained from any of the multiple orientation measurement units 20. θ，k σ φ，k If the azimuth measurement error information is not stored in the storage device 30, the error setting unit 42 consistently sets the azimuth measurement error (σ) attached to the azimuth measurement values ​​shown by the azimuth information read from the multiple azimuth measurement units 20 by the azimuth information reading unit 41. θ，k σ φ，k ).

[0058] The azimuth measurement value (θ) set by the error setting unit 42 k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The location information is read in by the directional information reading section 41.

[0059] Multiple orientation measuring units 20 output values ​​representing the orientation measurement error (σ). θ，k σ φ，k In the event of an error in the orientation measurement information, the multiple orientation measurement units 20 each function as an error setting unit 42.

[0060] The azimuth information reading unit 41 reads the time when each of the multiple azimuth sensors 10 observes the target 100, and the azimuth measurement values ​​(θ) accumulated in the storage device 30 representing the values ​​from the multiple azimuth measurement units 20 are read from the storage device 30. k φ k The azimuth information and the azimuth measurement value (θ) k φ k The associated measurement time t in ) k and the position information of each orientation sensor 10 (x r k yr k z r k ( ) location information.

[0061] At the same time, the storage device 30 also reads the time when each of the multiple orientation sensors 10 observes the target 100, and the stored data representing the orientation measurement error (σ). θ，k σ φ，k The azimuth measurement error information or the azimuth measurement value (θ) consistently set by the error setting unit 42. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ).

[0062] The time-domain segmentation unit 43 reads the azimuth measurement values ​​(θ) from multiple azimuth measurement units 20 according to each pre-set time-domain segmentation by the azimuth information reading unit 41. k φ k The location information is provided. The set time domain is a short-term time domain.

[0063] The time domain segmentation unit 43 divides the time when the target 100 is observed by the multiple azimuth sensors 10 into multiple time domains in a time sequence, and divides the azimuth information read by the azimuth information reading unit 41 according to each segmented time domain.

[0064] The time-domain segmentation unit 43 divides the azimuth information into M segments according to a pre-set time interval. This azimuth information represents the position based on the measurement time t. k The azimuth measurement value (θ) received by the azimuth information reading unit 41 from the radio waves received by the multiple azimuth sensors 10 from the target 100 is stored in the storage device 30. k φ k ), azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), measurement time t k and the position information of each orientation sensor 10 (x r k y r k z r k ).

[0065] Let m be the time domain divided into M parts, and let t be the representative time of each time domain m. m .

[0066] Furthermore, in the time domain m, the initial azimuth measurement value number among the azimuth information and azimuth measurement error information of the radio waves received from the target 100 by each azimuth sensor 10 according to the time series is set as k. m，s， The final orientation measurement value is numbered k. m，e .

[0067] For example, it means representing the K azimuth measurements (θ) taken during the period when the azimuth sensor 101 observes the target 100. k φ k The azimuth information is divided into M azimuth information segments according to the time sequence and at equal time intervals. The initial azimuth information segment in the m-th segment is labeled k. m，s The bearing measurement value is numbered, and the final bearing information is labeled k. m，e The azimuth measurement value number.

[0068] Even if the azimuth measurement value (θ) is taken k φ k The position information (x) of the orientation sensor 101 is divided into M equal intervals according to the time sequence. 1 k y 1 k z 1 k ) and representing the orientation measurement error (σ) θ，k σ φ，k The azimuth measurement error information is also included with the corresponding azimuth measurement value (θ). k φ k ).

[0069] Currently, for the case where there are 4 of 10 orientation sensors, the following is used: Figure 2 This illustrates an example of how the time-domain segmentation unit 43 divides the azimuth information representing the azimuth measurement value into M time-domain segments m at equal intervals according to the time interval W1, in a time-series order.

[0070] exist Figure 2 In, θ k (s1) represents the azimuth measurement value θ based on the azimuth information of the radio waves received by the first azimuth sensor 101. k θ k (s4) represents the azimuth measurement value θ based on the azimuth information of the radio waves received by the fourth azimuth sensor 104. k , 1 to M represent the first time domain to the Mth time domain.

[0071] In addition, representatively, k 1，s The azimuth measurement value θ in time domain 1 represents the initial azimuth information based on the radio waves received by the first azimuth sensor 101.k k 1，e The azimuth measurement value θ in time domain 1 represents the final azimuth information based on the radio waves received by the fourth azimuth sensor 104. k k M，s The azimuth measurement value θ in the time domain M represents the initial azimuth information based on the fourth azimuth sensor 104. k k M，e The azimuth measurement value θ in the time domain M represents the final azimuth information based on the radio waves received by the fourth azimuth sensor 104. k .

[0072] Thus, the time-domain segmentation unit 43 will use the indicated azimuth measurement values ​​(θ) of the radio waves received by each orientation sensor 10. k φ k The azimuth information is divided into M time-domain segments m at equal intervals according to the time interval W1, based on the azimuth sensor 10 and each time-domain segment m, and then further divided into K azimuth measurements (θ) according to the time series. k φ k ).

[0073] Furthermore, in one time domain m for a single orientation sensor 10, such as Figure 2 As shown, multiple azimuth measurements (θ) can exist. k φ k In addition, the following time-domain m may also exist: no azimuth measurement value (θ) exists. k φ k ) or there is one azimuth measurement value (θ) k φ k ).

[0074] The outlier determination unit 44 determines the azimuth measurement value (θ) for each time domain segmentation m set by the time domain segmentation unit 43. k φ k ), determine the azimuth measurement value (θ) existing in each set time domain m. k φ k For anomalies, a reliability level is set based on the determined degree of anomaly, and reliability information is obtained.

[0075] The outlier determination unit 44 performs the following processing: It implements the azimuth measurement values ​​(θ) present in each time domain m as set by the time domain segmentation unit 43. k φ k Whether this indicates an outlier or not, reset the azimuth measurement value (θ) that was determined to be an outlier. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，kIn the measurement processing for calculating the positional values, the azimuth measurement value (θ) that is judged as an outlier is reduced. k φ k The reliability of θ is reduced, which decreases the accuracy of azimuth measurements that are deemed outliers. k φ k The reliability of trajectory estimation.

[0076] On the other hand, the outlier determination unit 44 will determine it as the azimuth measurement value (θ). k φ k The azimuth measurement (θ) is not an outlier. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k (Keep it as is) In the measurement processing for calculating the positional values, the azimuth measurement value (θ) that will be determined not to be an outlier will be... k φ k The reliability of the azimuth measurement (θ) will remain unchanged, and the azimuth measurement value (θ) that is judged to be an outlier will be treated as an outlier. k φ k The reliability of the trajectory estimation remains unchanged.

[0077] In addition, the measurement processing for calculating the measurement value can be done using generally known processing methods.

[0078] The outlier determination unit 44 processes each time domain m segmented by the time domain segmentation unit 43 as follows: it determines the azimuth measurement value (θ) present in the time domain m. k φ k Does θ represent the azimuth of target 100? k φ k When the azimuth of target 100 is not indicated, the azimuth measurement value (θ) will be used. k φ k If a value is identified as an outlier, the azimuth measurement value (θ) will be reset. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The result is a reduction in the azimuth measurement (θ) that was identified as an outlier. k φ k The reliability of trajectory estimation.

[0079] The outlier determination unit 44 will determine it as the azimuth measurement value (θ) k φ k The azimuth measurement (θ) is not an outlier. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，kKeep it as is, and for the azimuth measurements (θ) that are determined not to be outliers... k φ k The reliability of the trajectory estimation remains unchanged.

[0080] In short, the outlier determination unit 44 determines the azimuth measurement value (θ) present in each time domain m divided by the time domain segmentation unit 43. k φ k Whether a value represents an outlier, the azimuth measurement value (θ) that will be judged as an outlier. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to reduce the measured value for that azimuth (θ) k φ k The reliability setting for trajectory estimation will be determined by the azimuth measurement (θ) that is not an outlier. k φ k The azimuth measurement error included in the calculation is reset to be applied to the azimuth measurement value (θ). k φ k The reliability of the trajectory estimation remains unchanged.

[0081] The positioning processing unit 45 uses the azimuth measurement value (θ) shown in the azimuth information to measure each time domain m segmented by the time domain segmentation unit 43. k φ k The azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the outlier determination unit 44 is reset. θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k The measurement process is performed to calculate the measurement value, and the measurement value of target 100 in time domain m is obtained, thus obtaining the target position information representing the measurement value of target 100 in time domain m.

[0082] The positioning processing unit 45, according to each time domain m divided by the time domain segmentation unit 43, uses k, which represents the number of the initial azimuth measurement value, existing in time domain m. m，s k represents the number of the final azimuth measurement. m，e Multiple azimuth measurements (θ) determined by the azimuth measurement value number k between them. k φ k ), multiple orientation measurement errors (σ) θ，kσ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k ), calculate the measurement value x m The positional processing yields the representative time t, which represents the time domain m. m The measured value x of the state m .

[0083] In addition, calculate the measurement value x m The positioning process can be handled using generally known methods.

[0084] Currently, the method involves using 4 of N orientation sensors 10 as the positional values ​​x of target 100. M In the case of orientation sensor 10, using Figure 3 This indicates that the measured value x in the time domain M is obtained. M The concept is as follows. The number of orientation sensors 10 is not limited to 4.

[0085] exist Figure 3 In the diagram, S1 to S4 represent the first azimuth sensor 101 to the fourth azimuth sensor 104. Line segments d1 to d4 extending from each of the first azimuth sensor S1 to the fourth azimuth sensor S4 represent the azimuth measurement values ​​θ of each of the first azimuth sensor S1 to the fourth azimuth sensor 104 in the time domain M. k x M，true The location of target 100 in the time domain M is shown.

[0086] That is, the azimuth measurement value θ k In representing the truth value x M，true Or it can represent a value close to the true value x. M，true The value is normal.

[0087] exist Figure 3 In the diagram, line segment d4 represents the orientation of target 100 in the time domain M (true value x). M，true The normal value of line segment d1 is the line segment d1, which is an outlier due to the multipath environment and does not represent the orientation (position) of target 100.

[0088] exist Figure 3 In the state shown, the anomaly determination unit 44 will determine the azimuth measurement value (θ) of the radio wave received by the first azimuth sensor 101. k φ k The value was determined to be an outlier, and the azimuth measurement value (θ) that was determined to be an outlier was reset. k φ kThe orientation measurement error (σ) included in the document θ，k σ φ，k ), reducing the azimuth measurement value (θ) that was judged to be an outlier. k φ k The reliability of trajectory estimation.

[0089] On the other hand, Figure 3 In the state shown, the anomaly determination unit 44 will determine the orientation (θ) of the radio waves received by the fourth orientation sensor 104. k φ k The value was determined to be within the normal range, and the orientation measurement error (σ) was within the normal range. θ，k σ φ，k Keep it as is, and for the azimuth measurement value (θ) that is judged to be normal... k φ k The reliability of the trajectory estimation remains unchanged.

[0090] That is, the outlier determination unit 44 reduces the azimuth measurement value (θ) that is determined to be an outlier. k φ k The reliability of ).

[0091] As a result, the positioning processing unit 45 uses the azimuth measurement values ​​(θ) based on the radio waves received by the first azimuth sensor 101 to the fourth azimuth sensor 104 respectively. k φ k ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the outlier determination unit 44 is reset. θ，k σ φ，k The measurement process is performed to calculate the position value, thus obtaining the estimated position of target 100 in the time domain m, i.e., the measurement value x. m The time domain m represents time t. m The measurement value x of target 100 m Target location information.

[0092] In the position measurement process at this time, the position measurement value is calculated based on the position measurement value (θ) of the radio wave received by the first azimuth sensor 101. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The reliability of the position measurement is reduced, therefore, it is possible to suppress the position measurement value (θ) based on the radio wave received by the first azimuth sensor 101 in the position measurement process. k φ k The positional value x, representing the estimated position of target 100, is obtained by the positional processing unit 45 under the influence of the positional effect. m As a truth value close to the target of 100, xM，true The value of .

[0093] Thus, as Figure 2 As shown, the position processing unit 45 obtains the representative time t from time domain 1 to time domain M. m The measured value x1 ~ the measured value x M .

[0094] like Figure 3 As shown, the output unit 46 will display the representative time t from time domain 1 to time domain M representing the estimated position of the target 100 obtained by the positioning processing unit 45. m The measured value x1 ~ the measured value x M All of them are connected in time sequence and output to output device 50 as the estimated trajectory of target 100.

[0095] The output device 50 is a device that utilizes object trajectory information from the output unit 46, such as a display that shows the trajectory of an object based on the object trajectory information.

[0096] like Figure 1 As shown, the outlier determination unit 44 includes a azimuth information reading unit 44a, a position gate determination unit 44b, and a reliability setting unit 44c.

[0097] The azimuth information reading unit 44a reads in time sequence the azimuth measurement values ​​(θ) of the radio waves received from the target 100 by the multiple azimuth sensors 10 in each time domain m set by the time domain segmentation unit 43. k φ k The azimuth information, representing the azimuth measurement value (θ) shown by the azimuth information. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The information on the azimuth measurement error and the measurement time t) k and the position information (x) of the orientation sensor 10 r k y r k z r k ).

[0098] For ease of explanation, the azimuth information reading unit 44a is described as a functional unit. However, the position gate determination unit 44b directly uses the azimuth measurement value (θ) read by the azimuth information reading unit 41, which is divided by the time domain segmentation unit 43 according to the time sequence and according to each time domain m, representing the azimuth measurement value of the radio waves received from the target 100 by the multiple azimuth sensors 10. k φ k The azimuth information, representing the azimuth measurement value (θ) shown by the azimuth information.k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The information on the azimuth measurement error and the measurement time t) k和 Position information of orientation sensor 10 (x) r k y r k z r k ).

[0099] The positioning gate determination unit 44b performs the following processing: For each of the multiple azimuth sensors 10, a positioning gate based on temporary positioning values ​​and temporary positioning accuracy is used for each time domain m to determine the azimuth measurement value (θ) shown by the azimuth information existing in the time domain m. k φ k () outliers.

[0100] The azimuth measurement value (θ) is determined by the positioning gate determination unit 44b. k φ k The handling of outliers is as follows: Based on the azimuth measurements (θ) of the radio waves received by multiple azimuth sensors 10 in the time domain m... k φ k Does the location shown in the figure represent a single point, i.e., does it indicate a unique location, or does the azimuth measurement (θ) include a velocity component in the position measurement result? k φ k Whether the azimuth measurement (θ) is concentrated around the state vector consisting of position and velocity is used to determine the orientation measurement value. k φ k Is it an outlier or a normal value?

[0101] The position measurement value (θ) is determined in the position gate determination unit 44b. k φ k In determining whether a value is an outlier or a normal value, multiple azimuth sensors are selected from N azimuth sensors 10. The azimuth measurement values ​​(θ) of the multiple azimuth sensors in each sensor group are determined by the judgment. k φ k Whether or not it exists within the measurement gate, evaluate the reliability. According to the sensor group with high reliability, the magnitude of the variance between the temporary measurement values ​​is taken as the anomaly of the measurement in the time domain. The time domain in which the anomaly is continuously lower than the set value is taken as the normal time domain, and the time domain in which the anomaly is continuously longer than the set value is taken as the abnormal time domain.

[0102] Alternatively, the time domain with the lowest anomaly can be taken as the normal time domain.

[0103] Currently, for the case where 4 of the N orientation sensors 10 are selected as the orientation sensors 10 to determine the estimated position of the target 100, the following method is used: Figure 4 This explains the concept of determining whether a value is abnormal or normal by the positioning gate determination unit 44b. The number of orientation sensors 10 is not limited to four.

[0104] exist Figure 4 In the diagram, S1 to S4 represent the first azimuth sensor 101 to the fourth azimuth sensor 104. The dashed line segments d1 to d4 extending from each of the first azimuth sensor S1 to the fourth azimuth sensor S4 represent the azimuth measurement values ​​θ of each of the first azimuth sensor S1 to the fourth azimuth sensor 104 in the time domain m. k x m，true This shows the position of target 100 in the time domain m.

[0105] Furthermore, as a premise, let's assume that the azimuth measurements (θ) of the radio waves received by the first azimuth sensor S1 to the fourth azimuth sensor S4 are respectively... k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ) are set to be equal.

[0106] The azimuth measurement (θ) is based on the radio waves received by the first azimuth sensor S1 to the fourth azimuth sensor S4 respectively. k φ k The azimuth measurement error (σ) attached to the azimuth measurement value. θ，k σ φ，k ), measurement time t k and the position information of each orientation sensor 10 (x r k y r k z r k ), perform positioning processing to obtain the positioning value x of target 100. m .

[0107] Regarding the measurement value x m Deviation from the true value x M，true The distance, for example, the azimuth measurement θ represented by line segment d2. k The azimuth measurement value θ represented by line segment d3 k When the correlation is close to linear, the azimuth measurement θ represented by line segment d1 is equally trusted. k and the azimuth measurement value θ represented by line segment d4 k And calculate the measurement value x m Therefore, it deviates significantly from the true value x. M，true .

[0108] That is, the measured value x of target 100 m Because the azimuth measurement value θ represented by line segment d1 k The influence of x causes it to deviate from the true value of target 100. M，true .

[0109] On the other hand, when the trusted azimuth measurement value (θ) k φ k When the unique location or target 100 in the location set exists in the state vector, regarding Figure 4 The location of target 100 can be trusted to exist, represented by the azimuth measurement θ, which is expressed by line segment d2. k The azimuth measurement value θ represented by line segment d3 k and the azimuth measurement value θ represented by line segment d4 k The location of the set is the truth value x M，true The surrounding area.

[0110] Therefore, the azimuth measurement value θ represented by line segment d1 can be expressed as... k It was determined to be an outlier.

[0111] The positioning gate determination unit 44b resets the azimuth measurement value θ, represented by line segment d1, which was judged as an outlier, by reducing the contribution of the positioning processing in calculating the positioning value. k The accompanying orientation measurement error (σ) θ，k σ φ，k ).

[0112] Next, use Figure 5 This explains how the position gate determination unit 44b handles the determination and processing of abnormal values.

[0113] Currently, the explanation focuses on the case where four of the N orientation sensors 10 are selected as the orientation sensors 10 for determining the estimated position of the target 100 and for handling outlier values. The number of orientation sensors 10 is not limited to four.

[0114] exist Figure 5 In the diagram, S1 to S4 represent the first azimuth sensor 101 to the fourth azimuth sensor 104. The dashed line segments d1 to d4 extending from each of the first azimuth sensor S1 to the fourth azimuth sensor S4 represent the azimuth measurement values ​​θ of each of the first azimuth sensor S1 to the fourth azimuth sensor 104 in the time domain m. k .

[0115] The position determination unit 44b performs the following processing.

[0116] First, select a certain number (3 in this example) of azimuth sensors from the first azimuth sensor S1 to the fourth azimuth sensor S4 and group them. For example, group them into a group of azimuth sensors S1, S2, and S3 (group 1), a group of azimuth sensors S1, S2, and S4 (group 2), a group of azimuth sensors S1, S3, and S4 (group 3), and a group of azimuth sensors S2, S3, and S4 (group 4).

[0117] Second, in each of the first to fourth sensor groups, according to each time domain m divided by the time domain segmentation unit 43, the azimuth measurement value (θ) of the radio waves received by the azimuth sensor S in each group is calculated. k φ k Perform positional analysis on known calculated positional values ​​to estimate the temporary positional value x. m，{1，2，3} Temporary measurement value x m，{1，2，4} Temporary measurement value x m，{1，3，4} Temporary measurement value x m，{2，3，4} Temporary positioning accuracy R m，{1，2，3} Temporary positioning accuracy R m，{1，2，4} Temporary positioning accuracy R m，{1，3，4} Temporary positioning accuracy R m，{2，3，4} .

[0118] Third, in the sensor group, for each time domain m, the estimated corresponding temporary measurement value x is set. m，{1，2，3} Temporary measurement value x m，{1，2，4} Temporary measurement value x m，{1，3，4} Temporary measurement value x m，{2，3，4} The determined positioning gates Ga~Gd.

[0119] The azimuth measurement value (θ) of the radio waves received by the first azimuth sensor S1 to the fourth azimuth sensor S4 is determined according to each time domain m. k φ k Does it exist within the position gate Ga~Gd?

[0120] Fourth, perform the following position gate determination based on the azimuth measurement values: In each sensor group, according to each time domain m, based on the azimuth measurement value (θ) k φ k The determination of whether the position exists within the corresponding position gate Ga~Gd is used to evaluate the azimuth measurement values ​​(θ) of each sensor group. k φ k The reliability of the measurement result is a temporary measurement value x. m，Ai Under conditions of high reliability, that is, in the orientation measurement values ​​(θ) of each sensor group k φ kWhen the azimuth measurement exists within each position gate Ga~Gd, the consistency of the azimuth measurement is determined, and the consistent azimuth measurement value (θ) is used as the basis for determining the consistency. k φ k The value is determined to be normal, and the azimuth measurement value (θ) that is not consistent across all sensor groups will be considered as normal. k φ k This value is identified as an outlier.

[0121] In short, the positioning gate determination unit 44b selects a certain number (3 in one example) of positioning sensors from multiple positioning sensors S1 to S4 and groups them into multiple sensor groups. Within each sensor group, a temporary positioning value is estimated according to each time domain m segmented by the time domain division unit 43. Positioning gates G1 to G4, determined by the estimated temporary positioning values, are set. The positioning measurement value (θ) shown by the positioning information present in the time domain m is then determined. k φ k When the value exists within the pre-defined positioning gates G1 to G4, it is considered a normal value. This is based on the azimuth measurement value (θ) shown by the azimuth information present in the time domain m. k φ k If a value does not exist within any of the pre-defined position gates G1 to G4, it is considered an abnormal value.

[0122] The following explanation uses mathematical formulas to describe the position gate determination of the position gate determination unit 44b.

[0123] Currently, the number of orientation sensors 10 capable of observing target 100 is set to N, and the set of multiple numbered orientation sensors 10 used in the positioning in time domain m is set to U={1, 2, ..., N}. The measurement time t used to determine time domain m is... k Let S be the set of the azimuth measurement values ​​numbered k. m ={k m，s , ..., k m，e}

[0124] Azimuth measurement value number k and measurement time t k The azimuth measurement value (θ) k φ k A one-to-one correspondence. k m，s The azimuth measurement value (θ) of the initial azimuth information in the time domain m k φ k Corresponding to k m，e The azimuth measurement value (θ) of the initial azimuth information in the time domain m k φ k )correspond.

[0125] The combination of selecting 3 orientation sensors 10 from set U is: N Type C3.

[0126] When the selected sensor group is set to A i (i=1, 2, ..., I) (I= N When C3), the combination of sensors is represented by the following formula (1).

[0127] A1={1,2,3}, A2={1,2,4},…,A I ={N-2, N-1, N}

[0128] ...(1)

[0129] Sensor group A i Let S be the set of azimuth measurement values ​​k that exist in the m-th time domain. m,Ai The set S is represented by the following equation (2) according to the time sequence. m,Ai The orientation measurement value is numbered k.

[0130] k=k （m,Ai），s ..., k （m,Ai），e ...(2)

[0131] That is, in the above formula (2), the azimuth measurement value number k is represented as k m,Ai Each azimuth measurement value.

[0132] When the orientation measurement value number k is represented by the above formula (2), in the set shown in the following formula (3), as shown in the following formula (4), the sets have overlapping orientation measurement value numbers.

[0133]

[0134]

[0135] In the positioning gate of the positioning gate determination unit 44b, by first using sensor group A i The azimuth measurement value (θ) k φ k ), and perform position processing on the known calculated position values ​​to obtain the representative time t shown in the following formula (5) for time domain m. m，Ai temporary measurement value x m，Ai and temporary positioning accuracy R m，Ai .

[0136]

[0137] In sensor group A i In the middle, the temporary measurement value x m，Ai Using the orientation measurement value (θ) existing in the time domain m k φ k ) and azimuth measurement (θ) k φ kThe orientation measurement error (σ) included in the document θ，k σ φ，k The position is determined by performing a measurement process.

[0138] In addition, in sensor group A i In the middle, the temporary positioning accuracy R m，Ai Using temporary measurement value x m，Ai and temporary measurement value x m，Ai The corresponding position and orientation measurement error (σ) of the orientation sensor S θ，k σ φ，k To estimate.

[0139] As a measurement method for measurement processing, methods such as weighted least squares estimation or MAP (Maximum a Posteriori) estimation can be used.

[0140] Furthermore, in state (measurement value) x m When the value contains not only position components but also velocity components, state x m It has a state vector with 6 components based on position and velocity in three-dimensional space. Alternatively, the following method shown in non-patent literature can be used: estimate the position and velocity of the target in MAP estimation based on the different azimuth measurement times and azimuth measurement errors of multiple sensors.

[0141] Temporary positioning accuracy R m，Ai For example, the following method described in non-patent literature can be used to estimate the observation error covariance with the BCRB (Bayesian Cramer-Rao Boundary) as the lower bound.

[0142] Non-patent literature: L.Badriasl, s. Arulampalam and A. Finn, “A Novel BatchBayesian WIV Estimator for Three-Dimensional TMA Using Bearing and ElevationMeasurements,” IEEE Transactions on Signal Processing, vol.66, no.4, pp.1023-1036, Feb.15, 2018.

[0143] The k-th orientation measurement value (θ) in time domain m k φ k The residual L) k，AiThe statistical distance is calculated using the measurement time t shown in equation (5) above. m，Ai The state transition matrix Φ shown in equation (7) k|（m，Ai） Converted to measurement time t k The temporary measurement value x obtained is shown in the following formula (6). k，Ai The temporary positioning accuracy R shown in equation (8) below k，Ai definition.

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150]

[0151] Additionally, the second term Qk on the right-hand side of equation (8) above is the driving noise covariance. Equation (9) above represents the difference between the predicted value and the determined azimuth value (θ) of the target object. k φ k The residual e) k，Ai Equation (10) above represents sensor group A. i The residual error covariance of DOA in the equation. H on the right-hand side of equation (10) above. k It is the partial differential matrix of the position component relative to the angle component. Equation (11) above represents the error covariance Σ of DOA. θφ，k Equation (12) above represents the k-th azimuth measurement value (θ) in the time domain m. k φ k The residual L) k，Ai The square of.

[0152] In the temporary measurement value x k，Ai Sensor group A used in the calculation i The azimuth measurement value (θ) k φ k The set S m，Ai The azimuth measurement value k = k of a certain proportion parameter β or higher. （m，Ai），S ..., k （m，Ai），e residual L k，Ai If the value falls within the threshold α1, all orientation measurements within that threshold are marked.

[0153] When in sensor group A i The azimuth measurement value (θ) is marked in the middle. k φ k Let the set of ) be T. m，Ai´ At that time, it can be said that sensor group A satisfies the marking condition. i´ temporary measurement value x k，Ai´ Since the azimuth measurements are concentrated, it is very likely that the target is 100, belonging to set T. m，Ai´ The azimuth measurement value (θ) k φ k This is likely the normal value for the target.

[0154] In equations (13) and (14) below, IF is a decision that returns 1 if the condition in parentheses is met and 0 if it is not met.

[0155]

[0156]

[0157] This process is repeated to determine the presence of all sensor groups A. i The azimuth measurement value (θ) is considered normal in the middle. k φ k The set T) m，Ai´ At this point, the union T belonging to the following equation (15) will be... m The azimuth measurement value (θ) k φ k ) is determined as a normal value in the position determination, and the azimuth measurement value (θ) in the time domain m of the complement T_m^C shown in the following formula (16) is determined. k φ k That is, the azimuth measurement value (θ) of the azimuth sensor that has never been judged as having a normal value in the entire sensor group. k φ k This value is identified as an outlier.

[0158]

[0159]

[0160] Furthermore, at this time, not all sensor groups A will be affected. i The azimuth measurement value (θ) that was judged to be normal at least once. k φ k ) is contained in set T m In addition to the above, it can also perform spatial clustering of temporary position values ​​to estimate which sensor group A is in the middle. i To represent the same target 100, thus determining set T. m .

[0161] The reliability setting unit 44c sets the azimuth measurement value (θ) that is determined to be an abnormal value by the positioning gate determination unit 44b. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The accuracy setting for multipath environments is configured to degrade the azimuth measurement error, thus reducing the azimuth measurement value (θ). k φ k The reliability parameter value (σ) of ) θ，err´ σ φ，err´ ), to obtain the parameter value (σ) θ，err´ σ φ，err´ As a reset orientation measurement error, it is used by the positioning processing unit 45.

[0162] The preferred method is to intentionally increase or decrease the azimuth measurement value (θ). k φ k The reliability parameter value (σ) of ) θ，err´ σ φ，err´ ).

[0163] The intentional increase mentioned here refers to the azimuth measurement value (θ) that will be judged as an outlier. k φ k σ in the orientation measurement error attached to ) θ，k The value is set to be greater than σ θ，k The value of σ is the largest among the parameter values. θ，err´ The value of .

[0164] In short, the azimuth measurement value (θ) will be judged as an outlier. k φ k The value of the orientation measurement error attached to the orientation measurement value is reset to a parameter value that is larger than the value of the orientation measurement error attached to the orientation measurement value that was judged to be normal.

[0165] That is, the parameter value is a parameter element determined by setting the reliability of the azimuth measurement value determined as an abnormal value relative to the azimuth measurement value determined as a normal value.

[0166] Furthermore, the reliability setting unit 44c will determine the azimuth measurement value (θ) as a normal value by the positioning gate determination unit 44b. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The position is kept unchanged and is used by the position processing unit 45 to ensure that the azimuth measurement value (θ) is not reduced. k φ k The reliability of the orientation measurement error has been reset and the original result has been maintained.

[0167] Next, based on the actions of the target trajectory estimation device, the following is used: Figure 6 This mainly explains the operation of the outlier determination unit 44.

[0168] The azimuth information reading unit 41 reads the azimuth measurement values ​​(θ) from the multiple azimuth measurement units 20, accumulated in the storage device 30, representing the time when the target 100 was observed by the multiple azimuth sensors 10. k φ k The azimuth information and the azimuth measurement value (θ) k φ k The measurement time t, which is attached to and related to the measurement in the ) k and the position information of each orientation sensor 10 (x r k y r k z r k ( ) location information.

[0169] Simultaneously, the storage device 30 reads the time when each of the multiple orientation sensors 10 observes the target 100, and stores the stored information representing the orientation measurement error (σ). θ，k σ φ，k The azimuth measurement error information or the azimuth measurement value (θ) set by the error setting unit 42. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ).

[0170] Next, the time-domain segmentation unit 43 determines the time based on the measured time t. k The azimuth information read in by the azimuth information reading unit 41 represents the azimuth measurement value (θ). k φ k ), azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), measurement time t k and the position information of each orientation sensor 10 (x r k y r k z r k The directional information is divided into M parts.

[0171] The azimuth measurement value (θ) of the target observed by the N sensors in each time domain m segmented by the time domain segmentation unit 43. k φ k ),according to Figure 6Steps ST1 to ST14 are shown, in which the outlier determination unit 44 determines whether the value is an outlier.

[0172] In step ST1, the azimuth information reading unit 44a reads the azimuth measurement values ​​(θ) of the time when the target was observed by the N sensors, which are divided according to the time domain m. k φ k ), azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k ).

[0173] Step ST1 is the step of reading the azimuth information of each time domain m, where the positioning gate determination unit 44b obtains the azimuth measurement values ​​(θ) divided into M units by the time domain segmentation unit 43. k φ k Steps such as )

[0174] Steps ST2 to ST14 are steps performed by the position gate determination unit 44b.

[0175] In step ST2, a certain number of azimuth sensors 10 are selected from the plurality of azimuth sensors 10 and grouped into multiple sensor groups. For example, from 4 azimuth sensors, a sensor group is formed with 3 azimuth sensors 10 as one group. Step ST2 is the step of selecting sensor groups.

[0176] In step ST3, the positional measurements (θ) are determined according to each time domain m in the assembled sensor group. k φ k Step ST3 involves performing measurement processing on the known calculated position values ​​to calculate the temporary position values. Step ST3 is the step of calculating the temporary position values ​​in each time domain m within each group.

[0177] For example, when grouping four azimuth sensors into groups of three (10 sensors each) into groups 1 through 4, in the time domain m, the temporary position value x for group 1 is calculated. m，{1，2，3} For the second group, calculate the temporary measurement value x. m，{1，2，4} For the third group, calculate the temporary measurement value x. m，{1，3，4} For the fourth group, calculate the temporary measurement value x. m，{2，3，4} .

[0178] The number of temporary position values ​​is calculated by multiplying the number of time domains by the number of sensor groups.

[0179] In step ST4, the temporary positioning accuracy attached to all the temporary positioning values ​​calculated in step ST3 is estimated. Step ST4 is the step of estimating the temporary positioning accuracy attached to the temporary positioning values.

[0180] The estimation of temporary positioning accuracy uses the temporary positioning value and the position information (x) of the azimuth sensor 10 corresponding to the temporary positioning value. r k y r k z r k ) and orientation measurement error (σ θ，k σ φ，k It was estimated using ).

[0181] For example, regarding the temporary measurement value x m，{1，2，3} Estimated temporary positioning accuracy R m，{1，2，3} For the temporary measurement value x m，{1，2，4} Estimated temporary positioning accuracy R m，{1，2，4} For the temporary measurement value x m，{1，3，4} Estimated temporary positioning accuracy R m，{1，3，4} For the temporary measurement value x m，{2，3，4 Estimated temporary positioning accuracy R m，{2，3，4} .

[0182] In step ST5, a positioning gate is set for all temporary positioning values ​​calculated in step ST3. Step ST5 is the step of generating the positioning gate.

[0183] The setting of the positioning gate is generated based on the temporary positioning value and the temporary positioning accuracy.

[0184] For example, regarding the temporary measurement value x m，{1，2，3} Set the position gate Ga for the temporary position value x. m，{1，2，4} Set the position gate Gb for the temporary position value x m，{1，3，4} Set the position gate Gc for the temporary position value x m，{2，3，4 Set the position gate Gd.

[0185] In step ST6, select the azimuth measurement value (θ) k φ k In step ST7, the selected azimuth measurement value (θ) is determined. k φ k Whether to enter the corresponding door. Step ST7 is the determination step for the orientation measurement value.

[0186] That is, given the existence of a selected azimuth measurement value (θ) k φ k The selected azimuth measurement value (θ) is determined in the time domain. k φ k ) Whether to enter the position gate G for the configured sensor group. If it is determined to enter, proceed to step ST8; if it is determined not to enter, proceed to step ST9.

[0187] In step ST8, the azimuth measurement value (θ) determined in step ST7 to be entering the positioning gate G is... k φ k Candidates are set to normal values.

[0188] In step ST9, it is determined whether all azimuth measurements (θ) were selected in step ST6. k φ k If not all are selected, return to step ST6 and repeat steps ST7 and ST8.

[0189] In step ST9, after selecting all azimuth measurements (θ) k φ k In the case of ), proceed to step ST10.

[0190] Thus, for all azimuth measurements (θ) read in step ST1 k φ k ), and determine whether to enter the position gate G.

[0191] In step ST10, the azimuth measurement value (θ) is determined for each time domain m in the assembled sensor group. k φ k A significant proportion of these values ​​entered the positioning gate G, namely the azimuth measurement values ​​(θ) selected in step ST8. k φ k The proportion of candidates set to the normal value is determined. When it is determined that a large proportion of candidates have entered the position gate G, the process proceeds to step ST11. When it is determined that no candidates have entered the position gate G, the process proceeds to step ST12.

[0192] In step ST11, when in step ST10, in the assembled sensor group, the azimuth measurement value (θ) k φ k When a relatively large proportion of the data enters the positioning gate G, the measured positioning value (θ) in this sensor group... k φ k () From a candidate of a normal value to a normal value.

[0193] In step ST12, the position measurement value (θ) measured in step ST11 is determined. k φ kIf a value is determined to be normal by the sensor group more than once, proceed to step ST13; if a value is determined to be normal by the sensor group more than once, proceed to step ST14.

[0194] In step ST13, the measured position values ​​(θ) that were determined to be normal values ​​of the sensor group more than once in step ST12 are... k φ k ) is set as the normal value in the outlier determination unit 44.

[0195] In step ST14, the measured value (θ) that was not determined to be a normal value of the sensor group even once in step ST12 is... k φ k ) is set as an outlier in the outlier determination unit 44.

[0196] Thus, for all azimuth measurements (θ) read in step ST1 k φ k Then, determine whether it is a normal value or an outlier.

[0197] In steps ST10 to ST14, in short, the selected azimuth measurement value (θ) in the time domain m is determined. k φ k The number of positions entering the positioning gate G is the selected azimuth measurement value (θ). k φ k The residual L) k，Ai Whether the number of items entering the threshold α1 is greater than or equal to the proportional parameter β.

[0198] In other words, determine the selected azimuth measurement value (θ) k φ k Whether it is concentrated at a single point, i.e., whether it represents a unique position, or whether it is concentrated around the state vector consisting of position and velocity.

[0199] Based on the results, determine whether the value is normal or outlier.

[0200] In step 15, the reliability setting unit 44c sets the azimuth measurement value (θ) that was determined to be an anomaly in step 14. k φ k Increase the azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reducing the azimuth measurement value (θ) k φ k The reliability of ).

[0201] For example, the orientation measurement error (σ) θ，kσ φ，k The setting was configured to reduce the azimuth measurement value (θ). k φ k The reliability parameter value (σ) of ) θ，err´ σ φ，err´ ).

[0202] Next, the positioning processing unit 45 uses the azimuth measurement value (θ) read by the azimuth information reading unit 41 according to each time domain m. k φ k ), azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the outlier determination unit 44 is reset. θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k The measurement process is performed to calculate the measurement value, resulting in the measurement value x for target 100 in the time domain m. m This yields the measured value x representing the position of target 100 in the time domain m. m Target location information.

[0203] Output unit 46 will represent the estimated position of target 100 obtained by position processing unit 45 at a time t from time domain 1 to time domain M. m The measured value x1 ~ the measured value x M All of them are connected in time sequence and output to output device 50 as the estimated trajectory of target 100.

[0204] The target trajectory estimation device is implemented through computer hardware architecture, such as... Figure 7 As shown, it has a processor 401, a memory 402, an input interface 403 and an output interface 404, and these components are interconnected via a bus 405.

[0205] The memory 402 has storage devices such as large-capacity semiconductor memory (RAM: Random Access Memory), hard disk devices or SSD devices, and non-volatile recording devices (ROM: Read Only Memory).

[0206] The processor 401 controls and manages the memory 402, the input interface 403, and the output interface 404.

[0207] The processor 401 temporarily saves the program stored in the ROM of the memory 402 to the RAM, and performs the process of generating the estimated trajectory of the target 100 according to the software / program saved in the RAM.

[0208] Input interface 403 is equivalent to Figure 1 The orientation information reading unit 41 shown, and the output interface 404 are equivalent to Figure 1 The output section 46 is shown.

[0209] Figure 1 The functions of the time domain segmentation unit 43, the outlier determination unit 44, the position processing unit 45, and the output unit 46 shown are implemented by the processor 401 loading the program stored in the memory 402 and performing actions according to the loaded program.

[0210] The target trajectory estimation method of the target trajectory estimation device, including steps ST1 to ST14, is performed by the processor 401 according to the program stored in the memory 402.

[0211] That is, the program stored in memory 402 has the following process: the time when the multiple azimuth sensors 10, which receive the arriving radio wave from the target 100, observe the target 100 is divided into multiple time domains m in a time sequence; and the azimuth measurement value (θ) obtained by the multiple azimuth sensors 10 is divided into each time domain m. k φ k The azimuth information is obtained; a certain number of azimuth sensors 10 are selected from multiple azimuth sensors 10 and grouped into multiple sensor groups; in each of the multiple sensor groups, the temporary position value x is estimated according to each time domain m segment. m，Ai Set the temporary measurement value x obtained from this. m，Ai Given a defined positioning gate G, when the azimuth measurement value (θ) is shown by the azimuth information existing in the time domain m... k φ k When the azimuth measurement value (θ) exists within the set positioning gate G, it is determined to be a normal value. k φ k If a value does not exist within any of the set position gates G, it is considered an outlier; the azimuth measurement value (θ) that is considered an outlier is... k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to reduce the measured value for that azimuth (θ) k φ k The reliability setting of trajectory estimation; according to the segmentation into each time domain m, the azimuth measurement value (θ) shown by the azimuth information existing in each time domain m is used. k φk ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the document has been reset. θ，k σ φ，k It estimates the position values ​​in each time domain m for target 100; and outputs object trajectory information that connects the estimated position values ​​in each time domain m in a time series to represent the trajectory of target 100.

[0212] As described above, the target trajectory estimation device 40 of Embodiment 1 includes: a time-domain segmentation unit 43, which divides the time when each of the multiple azimuth sensors 10 that receive the arriving radio wave from the target 100 observes the target 100 into multiple time domains m in a time-series order, and divides the azimuth measurement value (θ) obtained by each of the multiple azimuth sensors 10 into each of the divided time domains m. k φ k The azimuth information; the outlier determination unit 44, which determines the azimuth measurement value (θ) shown by the azimuth information in each time domain m according to the segmentation into each time domain m. k φ k Whether a value represents an outlier, the azimuth measurement value (θ) that will be judged as an outlier. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to reduce the measured value for that azimuth (θ) k φ k The reliability setting for trajectory estimation will be determined by the azimuth measurement error (σ) attached to the azimuth measurement values ​​that are not outliers. θ，k σ φ，k ), reset to use the measured value (θ) for this azimuth. k φ k The reliability of the trajectory estimation remains unchanged; the positioning processing unit 45 uses the azimuth measurement value (θ) shown by the azimuth information present in each time domain m according to each time domain m. k φ k ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the document has been reset. θ，k σ φ，k The positioning sensor 10 estimates the position values ​​in each time domain m for the target 100; and the output unit 46 outputs object trajectory information representing the trajectory of the target 100 by connecting the position values ​​in each time domain m estimated by the positioning processing unit 45 in a time sequence. Therefore, even if the positioning sensor 10 is in a multipath environment, it can estimate the position values ​​of the target 100 with high accuracy and perform trajectory estimation that is closer to the actual trajectory of the target 100.

[0213] In the target trajectory estimation device 40 of Embodiment 1, the azimuth measurement value (θ) in the outlier determination unit 44 k φ k In determining whether a value indicates an outlier, a certain number of azimuth sensors 10 are selected from multiple azimuth sensors 10 and grouped into multiple sensor groups. Within each sensor group, the azimuth measurement value (θ) shown by the azimuth information present in each time domain m is used according to the segmentation into which the azimuth is determined. k φ k ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The measurement process is performed to obtain the measurement value x from the sensor group targeting target 100. m Using the measurement value x m , and the measured value x m The corresponding position and orientation measurement error (σ) of the orientation sensor 10 θ，k σ φ，k Estimate the temporary positioning accuracy R of the sensor group. m By determining the azimuth measurement value (θ) shown by the azimuth information present in the time domain m. k φ k Does it exist in relation to the temporary measurement value x? m，Ai and temporary positioning accuracy R m，Ai Within a defined positioning gate G, evaluate the azimuth measurement value (θ). k φ k The reliability of the azimuth measurement (θ) is evaluated, and the azimuth measurement with high reliability is selected. k φ k The value θ is determined to be normal, and the azimuth measurement value with low reliability in this evaluation is considered to be normal. k φ k The value was determined to be an outlier; therefore, the azimuth measurement value (θ) was used. k φ k The consistent positioning gate G is set to the azimuth measurement value (θ). k φ k The reliability of the azimuth measurement (θ) is thus improved, allowing for the calculation of values ​​with reduced anomalies. k φ k The measured value of the contribution of ).

[0214] In addition, it is possible to determine the output of the normal azimuth measurement value (θ). k φ k The orientation sensor 10.

[0215] Furthermore, in the determination based on the position gate G, the position value x is calculated for each sensor group. m Therefore, outlier determination can be performed based on the temporary positioning accuracy R. m A defined positioning gate G is used to quantitatively determine the azimuth measurement value (θ). k φ k The anomaly was detected. Therefore, an azimuth measurement (θ) was established using multiple azimuth sensors 10. k φ k The positioning processing performed is robust to outliers generated in multipath environments, and it can still provide reliable azimuth measurements (θ) even in multipath environments. k φ k High-precision positioning processing.

[0216] In the target trajectory estimation device 40 of Embodiment 1, the outlier determination unit 44 includes a positioning gate determination unit 44b, which selects a certain number of positioning sensors 10 from a plurality of positioning sensors 10 and groups them into a plurality of sensor groups, and estimates a temporary positioning value x in each of the plurality of sensor groups according to each time domain m divided into segments. m，Ai Set the temporary measurement value x obtained from this. m，Ai Given a defined positioning gate G, when the azimuth measurement value (θ) is shown by the azimuth information existing in the time domain m... k φ k When the azimuth measurement value (θ) exists within the set positioning gate G, it is determined to be a normal value. k φ k If a value is not present in any of the set position gates G, it is determined to be an anomaly; and the reliability setting unit 44c determines the azimuth measurement value (θ) that is determined to be an anomaly. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to reduce the measured value for that azimuth (θ) k φ k The reliability of trajectory estimation is set, and the positioning processing unit 45 uses the azimuth measurement value (θ) shown by the azimuth information present in each time domain m to divide the time domain into segments. k φ k ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the document has been reset. θ，k σ φ，kThe positioning processing unit 46 estimates the position values ​​in each time domain m for the target 100, and outputs object trajectory information that represents the trajectory of the target 100 by connecting the estimated position values ​​in each time domain m in a time sequence. Therefore, the positioning processing unit 45 uses the azimuth measurement value (θ) that is determined to be an outlier. k φ k The accompanying reliability reduction due to the orientation measurement error (σ) θ，k σ φ，k The orientation sensor 10 performs positioning processing, so even in a multipath environment, it can accurately estimate the position value of the target 100 and perform trajectory estimation that is closer to the actual trajectory of the target 100.

[0217] In addition, it is possible to use the temporary measurement value x m，Ai A defined positioning gate G is used to quantitatively determine the azimuth measurement value (θ). k φ k ) abnormality.

[0218] Implementation method 2.

[0219] use Figures 8-10 This describes the target trajectory estimation device of Implementation Method 2.

[0220] The target trajectory estimation device of Embodiment 2 and the target trajectory estimation device of Embodiment 1 are divided into time domains m, and the target trajectory estimation device of Embodiment 2 is used to estimate the azimuth measurement values ​​(θ) obtained by multiple azimuth sensors 10. k φ k Compared to processing the azimuth information sequentially according to time series, the difference lies in that, during the period when multiple azimuth sensors 10 observe the target 100 (hereinafter referred to as the target observation time), the azimuth measurement values ​​(θ) obtained by multiple azimuth sensors 10 are acquired simultaneously. k φ k The estimated trajectory is processed in batches, and other aspects are the same.

[0221] In addition, Figures 8-10 In, with Figures 1-7 The same labels in the text indicate the same or equivalent parts.

[0222] Figure 8 This is a block diagram showing the general structure of a target trajectory estimation system including the target trajectory estimation device of Embodiment 2.

[0223] like Figure 8 As shown, the target trajectory estimation system has multiple orientation sensors 101 to 101. N Multiple orientation measuring units 201-20 N Storage device 30, target trajectory estimation device 40A, and output device 50.

[0224] Multiple orientation sensors 101-10 N Multiple orientation measuring units 201-20 N The storage device 30 and the output device 50, along with the plurality of orientation sensors 101 to 102 described in Embodiment 1, N Multiple orientation measuring units 201-20 N Since storage device 30 and output device 50 are the same, their descriptions are omitted.

[0225] The target trajectory estimation device 40A has an orientation information reading unit 41, an error setting unit 42, a time domain segmentation unit 43, an outlier determination unit 44A, a position processing unit 45, an accuracy estimation unit 47, a tracking processing unit 48, and an output unit 46.

[0226] The orientation information reading unit 41 and the time domain segmentation unit 43 are the same as those described in Embodiment 1, therefore, the description is omitted.

[0227] The outlier determination unit 44A includes a azimuth information reading unit 44a, a position gate determination unit 44b, a reliability setting unit 44c, a time domain selection unit 44d, a prediction information reading unit 44e, and a prediction gate determination unit 44f.

[0228] The orientation information reading unit 44a, the positioning gate determination unit 44b, and the reliability setting unit 44c are essentially the same as the target trajectory estimation device of Embodiment 1.

[0229] Regarding the position gate determination unit 44b, an example described in Embodiment 1 will be briefly explained.

[0230] The positioning gate determination unit 44b selects a certain number of positioning sensors 10 from the plurality of positioning sensors 10 and groups them into multiple sensor groups. In each of the multiple sensor groups, a temporary positioning value x is estimated according to each time domain m divided by the time domain segmentation unit 43. m，Ai Set the estimated temporary measurement value x m，Ai Given a defined positioning gate G, the azimuth measurement value (θ) is shown by the azimuth information existing in the time domain m. k φ k When the value exists within the pre-defined positioning gate G, it is considered a normal value. This is based on the azimuth measurement value (θ) shown by the azimuth information present in the time domain m. k φ k If a value does not exist within any of the predefined position gates G, it is considered an outlier.

[0231] Regarding the reliability setting unit 44c, an example described in Embodiment 1 will be briefly explained.

[0232] The reliability setting unit 44c sets the azimuth measurement value (θ) that is determined to be an abnormal value by the positioning gate determination unit 44b. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), resetting to reduce the azimuth measurement value (θ) k φ k The reliability parameter value (σ) of ) θ，err´ σ φ，err´ The azimuth measurement value (θ) determined as normal by the positioning gate determination unit 44b will be... k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to not reduce the azimuth measurement value (θ) k φ k The reliability of the system remains high and is maintained.

[0233] In short, the positioning gate determination unit 44b and the reliability setting unit 44c use the positioning gate G to measure all azimuth measurements (θ) existing in the entire time domain of the target observation time. k φ k The position gate is used to determine whether the measured value is normal or abnormal. Let the azimuth measurement value (θ) be determined to be normal. k φ k The reliability of the azimuth measurement (θ) is high, and the azimuth measurement value that is judged as an outlier is also high. k φ k The reliability of ( ) is low.

[0234] In the position gate determination of the position gate G in the entire time domain using the target observation time, the time domain selection unit 44d calculates the sensor group A that is determined to have high reliability. i´ temporary measurement value x m，Ai´ The variances of each other are selected based on their transformation, choosing the normal time-domain m as m. s .

[0235] Select the time domain m s Set it as the time domain of the tracking starting point.

[0236] Select time domain m s The method is to select the time domain with the smallest variance or the time domain with the least variance.

[0237] Position processing unit 45, according to the selected tracking starting point in the time domain m s For each time domain m adjacent to the starting point, along the positive and negative directions in chronological order, use the azimuth measurement values ​​(θ) shown in the azimuth information. k φ k The azimuth measurement value (θ)k φ k The orientation measurement error (σ) included in the reliability setting unit 44c is reset. θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k ), calculate the measurement value x m The positioning processing yields the positioning value x for target 100 in the time domain m. m This yields the measured value x representing the position of target 100 in the time domain m. m Target location information.

[0238] The accuracy estimation unit 47 uses the azimuth measurement error (σ) reset by the reliability setting unit 44c in the outlier determination unit 44A for each time domain m. θ，k σ φ，k ), estimating the measurement value x estimated by the measurement processing unit 45 m Positioning accuracy R m .

[0239] The accuracy estimation unit 47, according to each time domain m, calculates based on the measured position value x. m Position information of orientation sensor 10 (x) r k y r k z r k ) and orientation measurement error (σ θ，k σ φ，k Estimated measurement value x m Positioning accuracy R m .

[0240] The tracking processing unit 48 uses the measurement value x estimated by the measurement processing unit 45. m and the measured value x m The corresponding positioning accuracy R estimated by the accuracy estimation unit 47 m For the measured value x m Filtering is performed to obtain smoothed values ​​and predicted values.

[0241] The tracking processing unit 48 follows the tracking starting point time domain m s For each time domain m adjacent to the starting point, use the measurement value x m and positioning accuracy R m The position value x is calculated using a Kalman filter. m Filtering.

[0242] The tracking processing unit 48 obtains the time domain m of the tracking starting point. s The measurement value x adjacent to the starting point m The smoothed value x is obtained by weighting and averaging the predicted values ​​based on the motion continuity of target 100. m|m The smoothing value x m|m The output becomes the target position representing the moment.

[0243] In tracking processing methods, for example, there exists a possibility of outputting arbitrary measurement times t in parallel. k The predicted value x k|m Kalman filter.

[0244] In implementation method 2, the time domain m at the tracking starting point is... s The smoothing value x is obtained by applying a Kalman filter in the positive time direction to the positional processing of the starting point in the direction of time. m|m For tracking the starting point in the time domain m s The smoothing value x is obtained by applying a Kalman filter in the reverse time direction to the positional processing of adjacent time-domain locations starting from the time-backward direction. m|m .

[0245] The Kalman filter for tracking processing works as follows: based on the predicted value at the previous time step and the observed value at the current time step, it estimates the smoothed value (weighted average) at the current time step and the predicted value at the next time step (calculated using the motion model based on the smoothed value at the current time step).

[0246] Here, in the time domain of the tracking starting point m s In this process, it is necessary to select a time domain with low anomaly, such as a position value close to the true value, which can be judged by the position gate to function normally.

[0247] In the position gate determination, the temporary position value x is calculated. m，Ai´ This is used to determine whether the orientation measurements of the orientation sensor are concentrated at a single point or a single state vector.

[0248] Therefore, to confirm whether the positioning gate is functioning correctly, it is only necessary to evaluate the temporary positioning values ​​x of all sensor groups with high reliability in each time domain m. m，Ai´ The variance between them is sufficient; the magnitude of this variance can be considered as the anomaly of the measurement in the time domain.

[0249] When evaluating the anomaly degree across all time domains, the tracking starting point time domain m s Based on the ease of tracking and processing, select the time domain with consistently low anomaly or the time domain with the lowest anomaly.

[0250] First, using mathematical formulas, we apply them to the time domain m of the tracking starting point. sThe following is an explanation of the time-reverse Kalman filter for tracking processing in the time domain m.

[0251] Regarding the motion model of the Kalman filter in the reverse time direction, x m|m+1 Set as the predicted value, and set x m+1|m+1 Set as a smoothing value, using the value from the representative time t m+1 to t m The state transition matrix Γ in the reverse direction of time m|m+1 To record.

[0252] Predicted value x m|m+1 The state transition matrix Γ is expressed by the following equation (17). m|m+1 It is expressed by the following formula (18).

[0253]

[0254]

[0255] W m+1 ~N(0, Q´) m|m+1 The driving noise is the driving error covariance Q' with respect to the reverse time direction. m|m+1 When positive drive noise is added to the velocity component, the position component changes in the negative direction. Therefore, it is necessary to assign a negative sign to the off-diagonal blocks, and to denote Σ. q Set it as the covariance matrix of the three-dimensional position components as shown in equation (19).

[0256]

[0257] The observation model becomes x m =x m，true +v m .

[0258] v m ~N(0, R) m ) is observation noise, which is equivalent to the estimation error of the position value.

[0259] The initial steps of the Kalman filter in the reverse time direction (m=m s In the time domain of tracking the starting point m s The measured value x ms and positioning accuracy R ms Directly becomes the smooth value x ms|ms (=x) ms ) and smoothing error covariance P ms|ms (=R) ms ).

[0260] In m≦m s In the case of -1, the prediction processing in the reverse time direction is represented by the following equations (20) and (21).

[0261] x m|m+1 =Γ m+1 x m+1|m+1 (20)

[0262]

[0263] The smoothing process is represented by the following equations (22) to (24).

[0264] K m =P m|m+1 +(P) m|m+1 +R m ) -1 (twenty two)

[0265] x m|m =x m|m+1 +K m (x) m -x m|m+1 )(twenty three)

[0266] P m|m =P m|m+1 -K m P m|m+1 (twenty four)

[0267] P m|m+1 It is the prediction error covariance, K m It is the Kalman gain.

[0268] The tracking processing unit 48 uses a Kalman filter in the reverse time direction to measure the position value x in the time domain m. m The predicted value x of time domain m predicted by the above equation (20) at time domain m+1. m|m+1 The obtained smooth value x m|m The representative time t, which represents the estimated position of target 100 in output unit 46. m The output of the tracking processing unit 48 is used as the time domain m of the tracking start point. s Previous measurement value x m The time domain m representing the tracking starting point is connected sequentially according to the time series. s The object trajectory information of the previously estimated trajectory of target 100 is obtained.

[0269] On the other hand, the prediction information obtained by the tracking processing unit 48 is read in by the prediction information reading unit 44e, and the prediction gate determination unit 44f performs a position determination (θ) for the adjacent time domain m based on the prediction information. k φ k Outlier determination is called prediction gate determination.

[0270] For ease of explanation, the prediction information reading unit 44e is described as a functional unit; however, the prediction gate determination unit 44f directly uses the prediction information obtained by the tracking processing unit 48.

[0271] The prediction gate decision unit 44f performs the prediction using the predicted value x. k|m+1 The center determines the prediction gate.

[0272] The prediction gate determination unit 44f sets the prediction value x by the tracking processing unit 48. k|m+1 and the positioning accuracy R estimated by the accuracy estimation unit 47 m The estimated prediction accuracy P k|m+1 A defined prediction gate, where the azimuth measurement (θ) is given by the azimuth information present in the time domain m. k φ k When a value exists within the set prediction gate, it is determined to be a normal value, and the azimuth measurement value (θ) shown by the azimuth information present in the time domain m is considered normal. k φ k If a value does not exist in any prediction gate, it is considered an outlier.

[0273] The following method is expressed by formula: the prediction gate determination unit 44f uses the tracking processing of the Kalman filter in the reverse time direction to perform prediction gate determination.

[0274] Regarding the k-th orientation measurement value (θ) in the time domain m k φ k The residual L) k´， By using the prediction error covariance P k|m+1 The defined statistical distance is used to calculate the state transition matrix Φ. k|（m，Ai） Converted to measurement time t k The predicted value x k|m+1， It is expressed by the following formula (25).

[0275] Prediction error covariance P k|m+1 It is expressed by the following formula (27).

[0276] Equation (26) represents the state transition matrix Φ. k|m+1 .

[0277]

[0278]

[0279]

[0280]

[0281]

[0282]

[0283] Equation (28) above represents the difference between the predicted value and the measured value of the orientation of the object to be judged (θ). k φ k The residual e) k .

[0284] Equation (29) above represents the residual error covariance of the predicted DOA.

[0285] Equation (30) above represents the k-th azimuth measurement value (θ) in the time domain m. k φ k The residual L) k´ The square of.

[0286] The prediction value x at the center of the prediction gate is used by the prediction gate determination unit 44f in the prediction gate determination. m|m+1 The temporary measurement value x at the center of the measurement gate used by the measurement gate determination unit 44b in the measurement gate determination. k，Ai Unlike other methods, this method can obtain the location value through positional measurement processing. The positional measurement value has been subjected to anomaly detection based on the positional gate, thus ensuring high reliability.

[0287] Therefore, the reliability setting unit 44c determines the reliability based on the residual L. k´ Does it exceed the threshold α2? Reset the orientation measurement error (σ) θ，k σ φ，k ).

[0288] That is, as shown in equations (31) and (32) below, the residual L in if satisfies k´ When the condition is below the threshold α2, the orientation measurement error σ θk σ reset to normal value θk´ The azimuth measurement error σ φk σ reset to normal value φk´ The residual L in if condition is satisfied k´ When the condition exceeds the threshold α2, the orientation measurement error σ will be... θk Reset σ to an outlier θ，err´ The azimuth measurement error σ φk Reset σ to an outlier φ，err´ .

[0289] The orientation measurement error that has been reset (σ) θ，k σ φ，k ) was used to trace the starting point in the time domain m s The position processing in the position processing unit 45 starting from the next time domain m.

[0290]

[0291]

[0292] On the other hand, the tracking processing unit 48 pairs the tracking starting point time domain m s The position value x is obtained by applying a Kalman filter in the positive time direction to the time domain tracking process adjacent to the starting point in the direction of time. m .

[0293] That is, the tracking processing unit 48 obtains the measurement value x in the adjacent time domain m. m The predicted value x in time domain m-1 and the predicted value in time domain m. m|m-1 The smoothed value x obtained by weighted averaging m|m .

[0294] Calculate the predicted value x m|m-1 and smooth value x m|m The tracking process is essentially the same as that of a Kalman filter in the reverse time direction, which is a typical Kalman filter in the forward time direction; therefore, the explanation is omitted.

[0295] The tracking processing unit 48 uses a Kalman filter in the positive time direction to pass the position value x in the adjacent time domain m. m The predicted value x in time domain m-1 and the predicted value in time domain m. m|m-1 The smoothed value x obtained by tracking processing m|m The representative time t, which represents the estimated position of target 100 in output unit 46. m The measured value x m As the starting point for tracking in the time domain m s Subsequent measurement value x m The time domain m representing the tracking starting point is connected sequentially according to the time series. s The object trajectory information of the estimated trajectory of target 100 is obtained.

[0296] Output 46 serves as the time-domain starting point m obtained by tracking processing unit 48 using a Kalman filter in the positive time direction. s The estimated trajectory of target 100 and the time-domain tracking starting point m obtained by tracking processing unit 48 using a Kalman filter in the reverse time direction are then compared. s The estimated trajectories of the foremost target 100 are connected to obtain the object trajectory information representing the estimated trajectory of the target in the entire time domain during the observation time of target 100.

[0297] Furthermore, the prediction gate determination performed by the prediction gate determination unit 44f also uses tracking processing of the Kalman filter in the positive time direction, and the prediction gate determination unit 44f performs prediction based on the predicted value x. k|m-1 The prediction gate is determined by the center.

[0298] With the predicted value x k|m-1 The prediction gate determination centered on the time-reverse Kalman filter is essentially the same as the prediction value x. k|m+1 Since the prediction gate determination is the same as that of the center, the explanation is omitted.

[0299] The reliability setting unit 44c, similar to the time-reverse Kalman filter reset, is based on the residual L obtained by the prediction gate determination unit 44f. k´ Does it exceed the threshold α2? Reset the orientation measurement error (σ) θ，k σ φ，k ).

[0300] The reset azimuth measurement error (σ) is obtained similarly to the prediction gate determination of the Kalman filter in the opposite time direction. θ，k σ φ，k ) was used to trace the starting point in the time domain m s The position processing in the position processing unit 45 starting from time domain m.

[0301] Next, regarding the operation of the target trajectory estimation device, the focus is mainly on the tracking processing, which differs from that in Implementation Method 1, using... Figure 9 and Figure 10 This explains the operation of the outlier detection unit 44A.

[0302] Similar to the operation description in Implementation 1, the process continues until... Figure 6 Step ST15 is shown.

[0303] Figure 6 Steps ST12 to ST15 shown are equivalent to Figure 9 Steps ST101 and ST102 are shown.

[0304] That is, the positioning gate determination unit 44b performs positioning gate determination according to each time domain m, and determines the azimuth measurement value (θ) in the entire time domain (hereinafter referred to as the entire time domain) of the target observation time. k φ k The reliability setting unit 44c determines whether the value is normal or abnormal, and resets the azimuth measurement values ​​(θ) in all time domains. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ).

[0305] Steps ST101 and ST102 are reliability reset steps as follows: the position gate determination unit 44b and the reliability setting unit 44c set all azimuth measurement values ​​(θ) existing in the entire time domain. k φ kThe position gate G is used to determine whether the value is normal or abnormal. Let the azimuth measurement value (θ) be the one determined to be normal. k φ k The reliability of the azimuth measurement value (θ) is high, and the azimuth measurement value that is judged as an outlier is assumed to be an outlier. k φ k The reliability of ( ) is low.

[0306] In step ST103, during the position gate determination using the position gate G in the entire time domain in the time domain selection unit 44d, the sensor group A determined to have high reliability is calculated. i´ temporary measurement value x m，Ai´ The variance between each other.

[0307] In step ST104, based on the variance transformation calculated in the time domain selection unit 44d, the normal time domain m is selected as the tracking start time domain m. s .

[0308] Here, it is necessary to select a time domain with low anomaly, such as a position value close to the true value, that can be determined by the position gate to function normally, as the tracking starting time domain m. s .

[0309] In the position gate determination, the temporary position value x is calculated. m，Ai´ This is used to determine whether the orientation measurements of the orientation sensor are concentrated at a single point or a single state vector.

[0310] Therefore, to confirm whether the positioning gate is functioning correctly, it is only necessary to evaluate the temporary positioning values ​​x of all sensor groups with high reliability in each time domain m. m，Ai´ The variance between them is sufficient; the magnitude of this variance can be considered as the anomaly of the measurement in the time domain.

[0311] When evaluating the anomaly degree across all time domains, the tracking starting point time domain m s Based on the ease of tracking and processing, select the time domain with consistently low anomaly or the time domain with the lowest anomaly.

[0312] In the selected tracking starting point time domain m s Taking the positional processing of the time domain adjacent to the starting point in the positive direction of the time interval as an example, then entering... Figure 9 Step ST105, in the time domain of tracking starting point m s Starting from the positional measurement process in the adjacent time domain in the reverse direction of time backtracking, we enter... Figure 10 Step ST205.

[0313] We can proceed to step ST105, in the time domain starting from the tracking point m. sAfter the processing from the starting point to time domain M is completed, proceed to step ST205. Alternatively, you can proceed to step ST205 to trace the starting point in time domain m. s After the processing from the starting point to time domain 1 is completed, proceed to step ST105.

[0314] Figure 9 Steps ST105 to ST113 shown are steps in which a Kalman filter is applied in the positive time direction.

[0315] In step ST105, firstly, the tracking start point time domain m is... s The time domain m is chosen as the starting point for position measurement processing.

[0316] In step ST106, the positioning processing unit 45 reads the azimuth measurement value (θ) in the time domain m. k φ k The azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the reliability setting unit 44c is reset. θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k In step ST107, the read-in information is used for position measurement processing to calculate the position measurement value x for target 100 in the time domain m. m .

[0317] In step ST108, the accuracy estimation unit 47 estimates the position value x in the time domain m. m Position information of orientation sensor 10 (x) r k y r k z r k ) and orientation measurement error (σ θ，k σ φ，k ), estimate the measurement value x m Positioning accuracy R m .

[0318] In step ST109, the tracking processing unit 48 performs time-positive filtering on the time domain m and calculates the smoothing value x. m|m and predicted value x m+1|m .

[0319] The calculated smooth value x m|mThe representative time t used to represent the estimated position of target 100 m The output of the tracking processing.

[0320] In step ST110, when time domain m is time domain M, the process ends; when time domain m is not time domain M, proceed to step ST111, make time domain m the next time domain m+1, and proceed to step ST112.

[0321] In step ST112, the prediction gate determination unit 44f processes the predicted value x read by the prediction information reading unit 44e. k|m-1 The prediction gate is determined by the center.

[0322] In the prediction gate determination of the prediction gate determination unit 44f in step ST112, the prediction gate determination is performed based on the predicted value x. k|m-1 The updated azimuth measurement (θ) in time domain m k φ k Outlier determination is called prediction gate determination.

[0323] In step ST113, the reliability setting unit 44c will determine the azimuth measurement value (θ) that is identified as an outlier in the prediction gate determination. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The orientation error (σ) was reset to reduce reliability. θ，k σ φ，k The azimuth measurement value (θ) that will be judged as a normal value k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k Reset to maintain a high level of reliability and proceed to step ST105.

[0324] In step ST105, the updated time domain m from step ST111 is selected, and steps ST106 to ST113 are repeated until time domain m becomes time domain M in step ST110.

[0325] Thus, we obtain the time domain m from the tracking starting point. s The representative time t between time domain M and time domain M m The measured value x m .

[0326] In the positioning processing of the positioning processing unit 45, the azimuth measurement value (θ) of the time domain m is obtained for the adjacent time domain m-1. k φ k The prediction gate determination is performed from the tracking start point in the time domain m. sThe orientation measurement error (σ) has been reset throughout the entire time domain m between time domain M and time domain M. θ，k σ φ，k ).

[0327] Figure 10 Steps ST205 to ST213 shown are steps that apply a time-reverse Kalman filter.

[0328] In step ST205, firstly, the tracking start point time domain m is... s The time domain m is chosen as the starting point for position measurement processing.

[0329] In step ST206, the positioning processing unit 45 reads the azimuth measurement value (θ) in the time domain m. k φ k The azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the reliability setting unit 44c is reset. θ，k σ φ，k ), measurement time t k and the position information (x) of the orientation sensor 10 r k y r k z r k In step ST207, the read-in information is used for position measurement processing to calculate the position measurement value x for target 100 in the time domain m. m .

[0330] In step ST208, the accuracy estimation unit 47 estimates the position value x in the time domain m. m Position information of orientation sensor 10 (x) r k y r k z r k ) and orientation measurement error (σ θ，k σ φ，k ), estimate the measurement value x m Positioning accuracy R m .

[0331] In step ST209, the tracking processing unit 48 performs time-reverse filtering on the time domain m and calculates the smoothing value x. m|m and predicted value x m-1|m Predicted value x m-1|m The smoothing value x can be obtained using the above formula (17). m|m The answer can be obtained using the above formula (23).

[0332] The calculated smooth value x m|m The representative time t used to represent the estimated position of target 100 m The output of the tracking processing.

[0333] In step ST210, when time domain m is time domain 1, the process ends; when time domain m is not time domain 1, proceed to step ST211, make time domain m the next time domain m-1, and proceed to step ST212.

[0334] In step ST212, the prediction gate determination unit 44f processes the predicted value x read by the prediction information reading unit 44e. k|m+1 The prediction gate is determined by the center.

[0335] In the prediction gate determination of the prediction gate determination unit 44f in step ST212, a prediction gate determination is performed based on the predicted value x. k|m+1 The updated azimuth measurement (θ) in time domain m k φ k Outlier determination is called prediction gate determination.

[0336] In step ST213, the reliability setting unit 44c will determine the azimuth measurement value (θ) that is identified as an outlier in the prediction gate determination. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k The orientation error (σ) was reset to reduce reliability. θ，k σ φ，k The azimuth measurement value (θ) that will be judged as a normal value k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k Reset to maintain a high level of reliability and proceed to step ST205.

[0337] In step ST205, the updated time domain m from step ST211 is selected, and steps ST206 to ST213 are repeated until time domain m becomes time domain 1 in step ST210.

[0338] Thus, we obtain the time domain from time domain 1 to the tracking start point time domain m. s The representative moment t m The measured value x m .

[0339] In the positioning processing of the positioning processing unit 45, the azimuth measurement value (θ) for the adjacent time domain m+1 is calculated for time domain m. k φ k The prediction gate determination is performed from the tracking start point in the time domain m. sThe orientation measurement error (σ) has been reset in all time domains m between time domain 1 and time domain 2. θ，k σ φ，k ).

[0340] Output unit 46 will represent the estimated position of target 100 obtained by tracking processing unit 48 in steps ST109 and ST209 at a time t from time domain 1 to time domain M. m The measured value x1 ~ the measured value x M All of them are connected in time sequence and output to output device 50 as the estimated trajectory of target 100.

[0341] The target trajectory estimation device is similar to the target trajectory estimation device in Embodiment 1, and uses... Figure 7 The computer hardware structure shown is used to implement this.

[0342] Input interface 403 is equivalent to Figure 7 The orientation information reading unit 41 shown, and the output interface 404 are equivalent to Figure 7 The output section 46 is shown.

[0343] Figure 7 The functions of the time domain segmentation unit 43, outlier determination unit 44A, position processing unit 45, accuracy estimation unit 47, tracking processing unit 48, and output unit 46 shown are implemented by the processor 401 loading the program stored in the memory 402 and performing actions according to the loaded program.

[0344] The target trajectory estimation method of the target trajectory estimation device, which includes steps ST1 to ST14, ST101 to ST113 and ST201 to ST213, is performed by the processor 401 according to the program stored in the memory 402.

[0345] That is, the program stored in memory 402 has the following process: the time when the multiple azimuth sensors 10, which receive the arriving radio wave from the target 100, observe the target 100 is divided into multiple time domains m in a time sequence; and the azimuth measurement value (θ) obtained by the multiple azimuth sensors 10 is divided into each time domain m. k φ k The azimuth information is obtained; a certain number of azimuth sensors 10 are selected from multiple azimuth sensors 10 and grouped into multiple sensor groups; in each of the multiple sensor groups, the temporary position value x is estimated according to each time domain m segment. m，Ai Set the temporary measurement value x obtained from this. m，Ai Given a defined positioning gate G, when the azimuth measurement value (θ) is shown by the azimuth information existing in the time domain m... k φ kWhen the azimuth measurement value (θ) exists within the set positioning gate G, it is determined to be a normal value. k φ k If a value does not exist within any of the set position gates G, it is considered an outlier; the azimuth measurement value (θ) that is considered an outlier is... k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), reset to reduce the measured value for that azimuth (θ) k φ k The reliability setting of trajectory estimation; according to the segmentation into each time domain m, the azimuth measurement value (θ) shown by the azimuth information existing in each time domain m is used. k φ k ) and the azimuth measurement (θ) k φ k The orientation measurement error (σ) included in the document has been reset. θ，k σ φ，k Estimate the position values ​​in each time domain m for target 100; output the object trajectory information representing the trajectory of target 100 by concatenating the estimated position values ​​in each time domain m according to the time series; in the process of determining normal or abnormal values, calculate the temporary position value x of the sensor group determined to have high reliability. m，Ai The variances of each other are used to select the tracking starting point m based on the calculated variance transformation. s Using a Kalman filter in the positive time direction, the starting point in the time domain m is tracked. s The estimated positional value in the time domain m divided from the starting point and the associated redefined azimuth error (σ) within that positional value. θ，k σ φ，k ), estimate the positioning accuracy for the desired positioning value; use a Kalman filter in the positive time direction to track the starting point in the time domain m. s Given the estimated position value and corresponding position accuracy in the time domain m divided from the starting point, the position value is obtained by combining the predicted value and the smoothed value; a tracking starting point time domain m is defined. s The positional values, consisting of smoothed values, are divided into time domains m starting from the positional value, and the prediction gate is determined based on the prediction accuracy estimated according to the positional accuracy corresponding to the positional value. The azimuth measurement value (θ) is shown by the azimuth information existing in the adjacent time domain m+1 in the direction through which time passes. k φ k When the value exists within the set prediction gate, it is determined to be a normal value. When the azimuth measurement value (θ) shown by the azimuth information existing in the time domain is... k φ kIf a value does not exist within any prediction gate, it is considered an outlier; the tracking starting point time domain m s The redefined orientation measurement error (σ) in the next time domain m+1, which is divided from the starting point. θ，k σ φ，k The azimuth measurement value (θ) that is considered an outlier because it does not exist in any prediction gate. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), and then set to reduce the measured value (θ) for that azimuth. k φ k The reliability setting of trajectory estimation; using a time-reverse Kalman filter to track the starting point in the time domain m s The estimated position values ​​in the time domain m divided from the starting point and the previous time domain m-1, and the redefined azimuth measurement error (σ) attached to those position values. θ，k σ φ，k ), estimate the positioning accuracy for the desired positioning value; use a Kalman filter in the reverse time direction to track the starting point in the time domain m. s Given the estimated position value and corresponding position accuracy in the time domain m divided from the starting point, the position value is obtained by combining the predicted value and the smoothed value; a tracking starting point time domain m is defined. s The positional values ​​in the time domain m, divided from the starting point, are composed of smoothed values, and the prediction gate is determined based on the prediction accuracy estimated according to the positional accuracy corresponding to the positional values. The positional measurement value (θ) is shown by the azimuth information existing in the adjacent time domain m-1 in the opposite direction. k φ k When the value exists within the set prediction gate, it is determined to be a normal value. When the azimuth measurement value (θ) shown by the azimuth information existing in the time domain is... k φ k If a value does not exist within any prediction gate, it is considered an outlier; the tracking starting point time domain m s The redefined orientation measurement error (σ) in the previous time domain m-1, divided from the starting point. θ，k σ φ，k The azimuth measurement value (θ) that is considered an outlier because it does not exist in any prediction gate. k φ k The orientation measurement error (σ) included in the document θ，k σ φ，k ), and then set to reduce the measured value (θ) for that azimuth. k φ k The reliability setting of trajectory estimation; according to the segmentation into each time domain m, the azimuth measurement value (θ) shown by the azimuth information existing in each time domain m is used.k φ k ) and reset the azimuth measurement value (θ) k φ k The orientation measurement error (σ) included in the document has been reset. θ，k σ φ，k The resulting orientation measurement error (σ) θ，k σ φ，k It estimates the position values ​​in each time domain m for the target; and outputs object trajectory information that represents the trajectory of the target 100 by connecting the position values ​​of each time domain m, which are composed of smooth values, in a time series.

[0346] As described above, the target trajectory estimation device 40A of Embodiment 2, in addition to having the same effects as the target trajectory estimation device 40 of Embodiment 1, also includes: an accuracy estimation unit 47, which uses the azimuth measurement error (σ) reset by the outlier determination unit 44A for each time domain m divided by the time domain segmentation unit 43. θ，k σ φ，k ), estimating the measurement value x estimated by the measurement processing unit 45 m Positioning accuracy R m ; and tracking processing unit 48, which uses the measurement value x estimated by measurement processing unit 45. m and the measured value x m The corresponding positioning accuracy R estimated by the accuracy estimation unit 47 m For the measured value x m After filtering, smoothed values ​​and predicted values ​​are obtained. The outlier determination unit 44A, based on the predicted value from the tracking processing unit 48, re-sets the previously reset azimuth measurement error (σ). θ，k σ φ，k Therefore, by identifying two different complementary outliers, the trajectory estimation of target 100 using highly reliable positioning values ​​can be performed with higher accuracy.

[0347] In particular, trajectory estimation can be performed even in multipath environments where trajectory estimation is difficult to perform and are located around the target.

[0348] Furthermore, it is possible to freely combine the various embodiments, or modify any structural elements of the various embodiments, or omit any structural elements in the various embodiments.

[0349] Industrial availability

[0350] The target trajectory estimation device disclosed herein is suitable for use in target trajectory estimation systems for estimating the trajectory of moving objects such as aircraft.

[0351] Label Explanation

[0352] 10, 101~10 N : and multiple orientation sensors; 201~20 N 30: Multiple orientation measurement unit; 40: Storage device; 41: Target trajectory estimation device; 42: Orientation information reading unit; 43: Error setting unit; 44: Time domain segmentation unit; 44: Outlier determination unit; 44b: Positioning gate determination unit; 44c: Reliability setting unit; 44f: Prediction gate determination unit; 45: Positioning processing unit; 46: Output unit; 47: Accuracy estimation unit; 48: Tracking processing unit; 50: Output device; 100: Target.

Claims

1. A target trajectory estimation device, wherein, The target trajectory estimation device has: The azimuth information reading unit reads azimuth information obtained by multiple azimuth sensors that receive arriving radio waves from the target, representing azimuth measurement values, and azimuth measurement error information representing azimuth measurement errors that are attached to or uniformly set in the azimuth measurement values ​​shown by the azimuth information. The time-domain segmentation unit divides the time when the target is observed by the multiple azimuth sensors into multiple time domains in a time sequence, and segments the azimuth information read by the azimuth information reading unit according to each of the segmented time domains; The outlier determination unit determines whether an azimuth measurement value represents an outlier based on whether the positions represented by the azimuth measurement values ​​shown in the azimuth information in each time domain are concentrated at a single point or around the periphery of a state vector composed of position and velocity. If the azimuth measurement error attached to the azimuth measurement value determined to be an outlier is reset to reduce the reliability of the trajectory estimation for that azimuth measurement value, the azimuth measurement error attached to the azimuth measurement value determined not to be an outlier is reset to keep the reliability of the trajectory estimation for that azimuth measurement value unchanged. The positioning processing unit estimates the positioning value for the target in each time domain according to each time domain divided by the time domain segmentation part, using the azimuth measurement value shown by the azimuth information present in each time domain and the azimuth measurement error that has been reset in the azimuth measurement value. as well as The output unit outputs object trajectory information, which is the trajectory of the target represented by connecting the time-domain measurement values ​​estimated by the positioning processing unit in a time sequence.

2. The target trajectory estimation device according to claim 1, wherein, The determination of whether the azimuth measurement value in the outlier determination unit represents an outlier is as follows: A certain number of azimuth sensors are selected from the plurality of azimuth sensors and grouped into multiple sensor groups. In each of the multiple sensor groups, according to each time domain divided by the time domain segmentation part, the azimuth measurement value shown by the azimuth information in each time domain and the azimuth measurement error attached to the azimuth measurement value are used for positioning processing to obtain a temporary positioning value in the sensor group for the target. Using the temporary positioning value, the position of the azimuth sensor corresponding to the temporary positioning value and the azimuth measurement error, the temporary positioning accuracy of the sensor group is estimated. By determining whether the azimuth measurement value shown by the azimuth information in the time domain is within the positioning gate determined by the temporary positioning value and the temporary positioning accuracy, the reliability of the azimuth measurement value is evaluated. The azimuth measurement value with high reliability is regarded as a normal value, and the azimuth measurement value with low reliability is regarded as an outlier.

3. The target trajectory estimation device according to claim 1, wherein, The setting in the outlier determination unit that reduces the reliability of trajectory estimation for an outlier azimuth measurement value by reducing the azimuth measurement error attached to the azimuth measurement value is to reset the value of the azimuth measurement error attached to the azimuth measurement value that is determined to be an outlier to a parameter value that is larger than the value of the azimuth measurement error attached to the azimuth measurement value that is determined not to be an outlier.

4. The target trajectory estimation device according to claim 2, wherein, The estimation of the temporary positioning accuracy of the sensor group in the outlier determination unit is performed as follows: using the position of the azimuth sensor in the sensor group corresponding to the temporary positioning value and the azimuth measurement error set by the error setting unit, the lower limit of the variance that the temporary positioning value can theoretically take (BCRB: Bayesian Cramer-Rao Boundary) is calculated.

5. The target trajectory estimation device according to claim 1, wherein, The outlier determination unit has: The positioning gate determination unit selects a certain number of positioning sensors from the plurality of positioning sensors and groups them into multiple sensor groups. In each of the multiple sensor groups, a temporary positioning value is estimated according to each time domain segmented by the time domain division part. A positioning gate determined by the obtained temporary positioning value is set. When the positioning measurement value shown by the positioning information existing in the time domain exists within the set positioning gate, it is determined to be a normal value. When the positioning measurement value shown by the positioning information existing in the time domain does not exist within any of the set positioning gates, it is determined to be an abnormal value. as well as The reliability setting unit resets the azimuth measurement error attached to the azimuth measurement value that is determined to be an anomaly by the positioning gate determination unit to reduce the reliability of trajectory estimation for that azimuth measurement value.

6. The target trajectory estimation device according to claim 5, wherein, The estimation of the temporary measurement value by the measurement gate determination unit is performed by any method, either a weighted least squares estimation method or a MAP (Maximum a Posteriori) estimation method.

7. The target trajectory estimation device according to claim 1, wherein, In the determination of whether the azimuth measurement value represents an anomaly in the anomaly determination unit, a certain number of azimuth sensors are selected from the plurality of azimuth sensors and grouped into multiple sensor groups. In each of the multiple sensor groups, a temporary azimuth value and a temporary azimuth accuracy are estimated according to each time domain segmented by the time domain segmentation part. The reliability is evaluated by determining whether the obtained temporary azimuth value is within the azimuth gate determined by the obtained temporary azimuth value and the temporary azimuth accuracy. For each sensor group with high reliability, the variance between the temporary azimuth values ​​is taken as the anomaly degree of the azimuth measurement in the time domain.

8. The target trajectory estimation device according to any one of claims 1 to 7, wherein, The target trajectory estimation device also has: The accuracy estimation unit estimates the accuracy of the position measurement value estimated by the position measurement processing unit according to each time domain divided by the time domain segmentation part, using the orientation measurement error reset by the outlier determination unit. as well as The tracking processing unit uses the measured position value estimated by the positioning processing unit and the corresponding positioning accuracy estimated by the accuracy estimation unit to filter the measured position value, thereby obtaining a smoothed value and a predicted value. The outlier determination unit uses the predicted value from the tracking processing unit to reset the previously reset orientation measurement error.

9. The target trajectory estimation device according to claim 8, wherein, The tracking processing unit selects a time domain containing a azimuth measurement value that is determined by the outlier determination unit to be non-outlier. It then uses a Kalman filter in the positive time direction, starting from the selected time domain (i.e., the tracking start time domain), to calculate the predicted and smoothed values ​​in the next time domain after the tracking start time domain. Finally, it uses a Kalman filter in the negative time direction, starting from the tracking start time domain, to calculate the predicted and smoothed values ​​in the previous time domain before the tracking start time domain. The smoothed value in each time domain estimated by the tracking processing unit and used by the output unit is used as the output of the target trajectory estimation device.

10. The target trajectory estimation device according to claim 9, wherein, In the selection of the tracking start time domain by the tracking processing unit, the time domain with small variance of the azimuth measurement value or the time domain with the smallest variance is selected.

11. The target trajectory estimation device according to claim 9, wherein, In the selection of the tracking starting point time domain by the tracking processing unit, a certain number of azimuth sensors are selected from the plurality of azimuth sensors and grouped into multiple sensor groups. In each of the multiple sensor groups, a temporary measurement value and a temporary measurement accuracy are estimated according to each time domain segmented by the time domain segmentation part. The reliability is evaluated by determining whether the obtained temporary measurement value is within the measurement gate determined by the obtained temporary measurement value and temporary measurement accuracy. For each sensor group with high reliability, the variance between the temporary measurement values ​​is taken as the anomaly degree of the measurement in the time domain. The time domain with an anomaly degree that is consistently lower than a set value or the time domain with the lowest anomaly degree is selected.

12. The target trajectory estimation device according to claim 9, wherein, The outlier determination unit re-sets the previously reset azimuth measurement error in the following manner: for each time domain divided by the time domain segmentation part, a prediction gate determined by the prediction value of the tracking processing unit and the prediction accuracy estimated by the accuracy estimation unit is set. When the azimuth measurement value shown by the azimuth information existing in the time domain is within the set prediction gate, it is determined to be a normal value. When the azimuth measurement value shown by the azimuth information existing in the time domain is not within any of the set prediction gates, it is determined to be an outlier value.

13. The target trajectory estimation device according to claim 5, wherein, The target trajectory estimation device also has: The accuracy estimation unit estimates the accuracy of the position measurement value estimated by the position processing unit according to each time domain divided by the time domain segmentation part, using the azimuth measurement error reset by the reliability setting unit in the outlier determination unit. as well as The tracking processing unit uses a Kalman filter to filter the position values ​​estimated by the position processing unit in each of the adjacent time domains divided by the time domain segmentation, along with the position accuracy estimated by the accuracy estimation unit, to obtain predicted and smoothed values. The outlier determination unit includes a prediction gate determination unit. This prediction gate determination unit sets a prediction gate determined by the prediction value from the tracking processing unit and the prediction accuracy estimated by the accuracy estimation unit. When the azimuth measurement value shown by the azimuth information existing in the time domain exists within the set prediction gate, it is determined to be a normal value. When the azimuth measurement value shown by the azimuth information existing in the time domain does not exist within any prediction gate, it is determined to be an outlier. The reliability setting unit in the outlier determination unit will re-set the azimuth measurement error attached to the azimuth measurement value that has been determined as an outlier by the prediction gate determination unit, which is already reset to a setting that reduces the reliability of trajectory estimation for that azimuth measurement value. The smoothed value in each time domain estimated by the tracking processing unit and used by the output unit is used as the output of the target trajectory estimation device.

14. A target trajectory estimation method, wherein, The target trajectory estimation method has the following steps: The time-domain segmentation unit divides the time when the multiple azimuth sensors that received the arriving radio wave from the target observed the target into multiple time domains in a time-series order, and segments the azimuth information representing the azimuth measurement value obtained by the multiple azimuth sensors according to each of the segmented time domains; The position gate determination unit in the outlier determination unit selects a certain number of position sensors from the plurality of position sensors and organizes them into multiple sensor groups. In each of the multiple sensor groups, a temporary position value is estimated according to each time domain segmented, and a position gate determined by the obtained temporary position value is set. When the position measurement value shown by the position information existing in the time domain exists within the set position gate, it is determined as a normal value. When the position measurement value shown by the position information existing in the time domain does not exist within any of the set position gates, it is determined as an outlier value. The reliability setting unit in the outlier determination unit will be determined to be an azimuth measurement error attached to the azimuth measurement value of the outlier value, and will be reset to reduce the reliability of the trajectory estimation for that azimuth measurement value. The positioning processing unit estimates the positioning values ​​for the target in each time domain according to each of the time domains divided into segments, using the azimuth measurement values ​​shown by the azimuth information present in each time domain and the azimuth measurement error that has been reset in the azimuth measurement values. as well as The output unit outputs object trajectory information, which is the trajectory of the target by connecting the estimated position values ​​of each time domain in a time sequence.

Images