JP6251930B2 - Position estimation system - Google Patents

Description

    The present invention relates to position estimation of a terminal using radio. In particular, it is intended to reduce deterioration in position estimation accuracy caused by an environment outside the line-of-sight (Non-Line Of Light, hereinafter referred to as “NLOS”).

  As a general position detection technique, there is GPS (Global Positioning System; hereinafter referred to as “GPS”). GPS is a position detection system that uses radio waves from artificial satellites, and can estimate the position with high accuracy outdoors, but the position estimation accuracy deteriorates because the line of sight is blocked by obstacles such as the roof indoors. Resulting in. Therefore, indoors, the position of the target node is estimated by receiving radio waves from the target node by a plurality of sensor nodes installed in the space and integrating information obtained using the sensor network. The target node position is estimated by the arrival time of radio waves (Time of Arrival, hereinafter referred to as “TOA”), the arrival time difference (Time Difference of Arrival, hereinafter referred to as “TDOA”), and the direction of arrival (Angle of Arrival, hereinafter referred to as “AOA”). ”), The received signal strength (Received Signal Strength; hereinafter referred to as“ RSS ”) and the like are measured at each sensor node. In addition, a hybrid system (Hybrid TOA / AOA, hereinafter referred to as “HTA”) that improves estimation accuracy by simultaneously measuring TOA and AOA has been devised (see, for example, Non-Patent Document 1).

  When performing position estimation, measurement errors always occur in the measured values. When the frequency used for position estimation is an ultra-wideband (Ultra-Wide-Band; hereinafter referred to as “UWB”), it is strongly influenced by multipath, and a slight bias is applied to the TOA measurement value (for example, Non-Patent Document 2). reference). In addition, the radio wave transmitted from the target node is obstructed by an obstacle, and the position estimation accuracy is greatly deteriorated by the non-line-of-sight (hereinafter referred to as “NLOS”) environment where the direct wave does not reach the sensor node. . In contrast to a line-of-sight (hereinafter referred to as “LOS”) environment in which a wave directly reaches the sensor node, a reflected wave or diffracted wave of a radio wave transmitted from the target node is received by the sensor node in the NLOS environment. The measurement error is greatly increased. A well-known model of the NLOS environment is GSDM (Gaussian Scatter Density model). This model is one of the single scattering models in which the radio wave transmitted from the target node is scattered once by the scatterer and then received by the sensor node. The position of the scatterer is a two-dimensional model centered on the target node. This model is assumed to exist according to a Gaussian distribution (see, for example, Non-Patent Document 3).

  Since the error due to the NLOS environment leads to a significant deterioration in the position estimation accuracy, the reduction due to the NLOS environment is indispensable for improving the position estimation performance. One of the most well-known NLOS environmental measurement value removal methods is the IMR (Iterative Minimum Residual) method (see, for example, Non-Patent Document 4). The IMR method is a method of repeatedly estimating a target position by variously combining measurement values, and selecting and removing measurement values with low accuracy by comparing residuals obtained from the estimated target positions. By repeating this operation, inaccurate measurement values are removed one by one. The IMR method is suitable for a sensor network that requires a small amount of calculation for discrimination of a sensor node in an NLOS environment (hereinafter referred to as “NLOS discrimination”) and requires power saving. In particular, position estimation using TOA Well applied to formulas.

S. Venkatraman and J. Caffery, Jr., “Hybrid TOA / AOA Techniques for Mobile Location in Non-Line-of-Sight Environments,” WCNC.2004, pp.274-278, Vol.1, 2004. B Alavi, K pahlavan, “Modeling of the TOA-based Distance Measurement Error Using UWB Indoor Radio Measurements,” IEEE Communications, vol.10, no.4, 2006. Ramakrishna Janaswamy, “Angle and Time of Arrival Statistics for the Gaussian Scatter Density Model,” Wireless Communications, IEEE Transactions on, Vol. 1, Issue 3, pp.488-497, 2002. X. Li, “An iterative NLOS mitigation algorithm for location estimation in sensor networks,“ in Oroceedings of the 15th IST Mobile Wireless Communications Summit, Myconos, Greece, June 2006.

  However, when the IMR method is applied to the HTA method, there is a problem that when there are a plurality of sensor nodes in an NLOS environment, discrimination errors increase and the position estimation accuracy deteriorates. According to the inventor's investigation, the target node position estimation accuracy deteriorates because the position estimation using the AOA greatly varies due to a large error caused by the NLOS environment. Therefore, in the IMR method, when the sensor node in the NLOS environment is determined by comparing the residual obtained from the estimated position using the AOA, the accuracy of the determination is deteriorated if there are two or more sensor nodes in the NLOS environment.

In order to solve the above-mentioned problem, the invention according to claim 1 receives the signal transmitted from the target node (2), and based on the received signal, the distance from the target node (2) to the own device A plurality of sensor nodes (3-i) that output a set of distance information (d _i ) that can be calculated and arrival direction information (θ _i ) that can calculate the direction of arrival of the signal, and processing means (4 And the processing means (4) obtains a set of distance information (d _i ) and direction-of-arrival information (θ _i ) output from each of the plurality of sensor nodes (3-i). Of the distance acquisition unit (S100) and all the distance information (d _i ) and arrival direction information (θ _i ) acquired by the acquisition unit (S100), only the distance information (d _i ) Pair used for position estimation of node (2) To include, with the object, select a large distance information error _{(d i),} a selected distance information first removing means for removing _{(d i)} from the subject (S105, S110), said first removal means After the distance information (d _i ) having a large error is removed from the target in (S105, S110), a plurality of arrivals paired with any one of the plurality of distance information (d _i ) remaining in the target A direction information (θ _i ) is added to the target, and a pair of distance information (d _i ) and direction-of-arrival information (θ _i ) having a large error is removed from the target using the target after the addition. The target calculated using the target after the pair of distance information (d _i ) and arrival direction information (θ _i ) having a large error is removed by the removing means (S115) and the second removing means (S115). Get the estimated position of the node And an estimated position acquisition means (S120).

In this way, in the first removal means (S105, S110) in advance, based on only the distance information (d _i ) out of the distance information (d _i ) and the arrival direction information (θ _i ), the NLOS environment There is a high possibility that the distance information (d _i ) of the sensor node located at is excluded from the target used for position estimation of the target node (2). Therefore, at the start of execution of the second removal means (S115) using both the distance information (d _i ) and the arrival direction information (θ _i ), the distance information (d _i ) and the arrival direction of the sensor node in the NLOS environment The number of pieces of information (θ _i ) is reduced compared to the beginning, and as a result, the accuracy of discrimination is improved.

It is a figure which shows the system model of the position estimation system using the sensor network in embodiment of this invention. It is a figure which shows the flow of the NLOS discrimination | determination process and position estimation in this embodiment. It is a figure which shows the flow of the subroutine which performs NLOS discrimination | determination using only a TOA measurement value during NLOS discrimination | determination processing. It is a figure which shows the flow of the subroutine which performs NLOS discrimination | determination using the measured value of both TOA and AOA during NLOS discrimination | determination processing. It is a figure of the sensor field which is an area | region where a sensor node and a target node exist in computer simulation. It is a figure which shows the result of a computer simulation, and is a figure which shows RMSE of the whole sensor field at the time of applying the IMR method in the HTA system of this embodiment. It is a figure which shows the result of a computer simulation, and is a figure which shows RMSE of the whole sensor field at the time of applying the IMR method in a TOA system as an existing method. It is a figure which shows RMSE of the whole sensor field at the time of applying an IMR method to an HTA system simply.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a system model of a position estimation system using a sensor network according to the present embodiment. A position estimation system using a sensor network includes a single or a plurality of target nodes 2 (single in FIG. 1) and a plurality of sensor nodes 3-i (i = 1, 2,..., N) in the sensor field 1 (FIG. 1 includes N = 9) and the central processing unit 4. The target node 2 has a function of transmitting a signal wirelessly and portability. Each sensor node 3-i has an arrival time t _i (i = 1, 2,..., N), which is a time from when a signal is transmitted from the target node 2 until it arrives at its own device, and its own signal. Has a function of measuring an angle of arrival θ _i (i = 1, 2,..., N) representing the direction of arrival of. As a configuration and method for measuring the arrival time t _i and the arrival angle θ _i , a known configuration and method used in the TOA method and the AOA method may be used.

The positions of these sensor nodes 3-i are normally fixed and are known in the central processing unit 4 of the position estimation system. Here, the arrival time t _i (i = 1, 2,..., N) measured at the sensor node 3-i is multiplied by the speed of light c at each sensor node 3-i, thereby obtaining the sensor node 3-i / target. The distance d _i between nodes 2 (i = 1, 2,..., N). Here, d _i is called a TOA measurement value, and θ _i is called an AOA measurement value. A set of TOA measurement values (corresponding to an example of distance information) and AOA measurement values (corresponding to an example of direction-of-arrival information) obtained at each sensor node 3-i is transmitted to the central processing unit using wired or wireless communication. Is done. The central processing unit 4 integrates the measurement value data of the set of the TOA measurement value and the AOA measurement value received from each sensor node 3-i, and estimates the target node 2 using the data of the plurality of sets of measurement values. Determine the position.

  Here, there is a possibility that a measurement value (TOA measurement value, AOA measurement value, or both) including a large error exists due to the influence of the NLOS environment among the plurality of sets of measurement values obtained. When the measurement value of the sensor node 3-i in the NLOS environment is used for position estimation as it is, the position estimation accuracy is greatly deteriorated. Therefore, it is necessary to determine which measurement value is in the NLOS environment. .

  FIG. 2 shows a flow of NLOS discrimination processing and position estimation realized by the central processing unit 4 executing a program in the present embodiment. 2, the details of the IMR method subroutine (step S105) using only the TOA measurement values are shown in FIG. 3, and the IMR method subroutine (step S115) using both the TOA measurement values and the AOA measurement values are shown in detail. As shown in FIG. In S100, a set of TOA measurement values and AOA measurement values obtained and transmitted by each sensor node 3-i is acquired and integrated. At this time, which sensor node 3 is used to store each set of measurement values is stored. In subsequent S105, NLOS discrimination is performed using the obtained plural sets of measurement values.

  In this embodiment, the IMR method, which is an NLOS discrimination method, is performed in two steps. In the first IMR method, all the TOA measurement values among the plurality of sets of obtained measurement values are obtained. Only and not any AOA measurements. This IMR method is the same as the general IMR method, and will be described below in accordance with the flow shown in FIG. In S200, the estimated position of the target node 2 is derived using all N TOA measurement values used for the measurement. The estimated position (x, y) can be derived from the following equation by using a least squares (LS) method.

However, (x _i , y _i ) is the position of the sensor node 3-i, N _d is the number of measurement values used for calculating the estimated position, and since all measurement values are used initially, N _d = N It is. Equation (1) can be obtained by using a solution search technique such as the well-known Newton-Raphson method. Also, a normalized residual η is derived from the following equation from the obtained estimated position.

In S200, the normalized residual ε calculated using the equation (2) is set as a normalized residual η. Here, is normalized with (1) to eliminate the influence of residual is the difference between the measurement value of the number represented by the formula, the measurement value of the number N _d of using. At this point, the target used for position estimation of the target node 2 includes only N TOA measurement values among a set of N sets of TOA measurement values and AOA measurement values.

Subsequent S205 is decreased by one the value of the TOA measurements number _{N d} used in estimated position calculation. Combinations of extracting _{N d} number of TOA measurements from _N d + 1 pieces of TOA measurements currently contained in the target to be used for position estimation of the target node 2 is +1 ways _N d. In subsequent S210, the _N by using each of _{N d} number of TOA measurements group d +1 ways _to calculate the estimated position of the +1 kinds _N d (x, y) and the normalized residual epsilon. As a result, N _d +1 estimated positions and normalized residuals can be derived. The estimated position and the normalized residual can be derived by using the equations (1) and (2). Among the obtained N _d +1 normalization residuals ε, a normalization residual ε ′ that is the minimum value and an estimated position (x ′, y ′) at that time are selected.

In subsequent S215, the two normalized residuals η and ε ′ are compared. (Η-ε') is not less minute value _{[delta] TOA} or set, and _N when _d is greater than 3 (S215: NO), the process proceeds to S218.

In S218, among the N _d +1 TOA measurement values to be used for position estimation of the target node 2, it corresponds to the TOA measurement value that was not used in the calculation of the normalized residual ε ′ calculated in the immediately preceding S210. The sensor node 3-i to be processed is regarded as the sensor node 3-i in the NLOS environment. Specifically, the TOA measurement value is selected, and the selected TOA measurement value is removed from the target used for position estimation of the target node 2. Subsequently, in step S220, the value is changed to η = ε ′, the number of measurement values used for calculating the estimated position is reduced by 1, and the process returns to S210.

  By repeating this, the TOA measurement value with low accuracy is removed one by one from the target used for position estimation of the target node 2 one by one.

In S215, (η-ε') If is smaller than the very small value _{[delta] TOA,} or _{N if d} becomes 3 or less (S215: YES), the process proceeds to S225, returns the _{N d} at that time, TOA The NLOS discrimination using only the measured value is terminated.

  Here, in the NLOS discrimination using only the TOA measurement value, by repeating step S218 a plurality of times, a plurality of NLOS measurement values (TOA measurement values under the NLOS environment) from the target used for position estimation of the target node 2 are obtained. As a result of the removal, when the number of measurement values to be used decreases, the probability of erroneous NLOS discrimination increases. Therefore, in the NLOS discrimination using only the TOA measurement values, the probability that the last removed TOA measurement value is erroneously removed even though it is not actually the measurement value corresponding to the sensor node 3-i in the NLOS environment. Is expensive.

  Therefore, the measurement value finally determined to correspond to the sensor node 3-i under the NLOS environment is left without being finally removed, and once again using the TOA measurement value (and the paired AOA measurement value). It is preferable (but not essential) to perform NLOS discrimination by the IMR method.

Therefore, in S110, the central processing unit 4 returns the TOA measurement value finally removed by the first IMR method (S105) to the target used for position estimation of the target node 2 in order to use it in the next NLOS discrimination. At the same time, _Nd is increased by one.

In subsequent S115, all AOA measurement values paired with any one of a plurality of TOA measurement values included in the target used for position estimation of the target node 2 are selected, and the selected AOA measurement value is added to the target. . Then, the IMR method is performed. The description will be made according to the flow shown in FIG. In S300, all AOA measurement values that are paired with any of the TOA measurement values included in the target are added as targets used for position estimation of the target node 2. Then, N _d sets of measurement values contained in the object to be used for position estimation of the target node 2 (TOA measurements, AOA measurements) derives the estimated position and normalized residual of the target node 2 using. The estimated position is expressed by the following equation using the LS method as in the first IMR method.

However, γ _{i, TOA} , γ _{i, AOA} are weights for the i-th TOA and AOA measurement values, respectively, and are determined by the reliability of each measurement value. This weight is usually the reciprocal of the error variance. Further, the normalized residual obtained from the estimated position is expressed by the following equation.

In S300, the normalized residual ε calculated using the equation (6) is set as the normalized residual ζ. Here, the residual represented by the equation (6) is normalized by the number of measured values Nd used to eliminate the influence of the difference in the number of measured values.

Subsequent S305 is decreased by one the value of the TOA measurements number _{N d} used in estimated position calculation. _N d + 1 set of measurements that are currently in the object to be used for position estimation of the target node 2 (TOA measurements, AOA measurements) combination to extract the _{N d} sets of TOA measurements from one +1 ways _N d . In subsequent S310, using each of the _{N d} sets of measurements group +1 ways this _N d (group having a plurality of sets of TOA measurements and AOA _{measurements),} the estimated position of the _N d +1 ways (x, y) and normalized residual ε are calculated. As a result, N _d +1 estimated positions and normalized residuals can be derived. The estimated position and the normalized residual can be derived by using equations (3) to (6). Among the obtained N _d +1 normalization residuals, the normalization residual ε ′ that is the smallest value and the estimated position (x ′, y ′) at that time are selected.

In subsequent S315, the normalized residuals are compared. When (ζ−ε ′) is smaller than the set minute value δ _AOA (S315: YES), the process proceeds to S320, and when it is the same or larger (S315: NO), the process proceeds to S325. In S320, (x, y) calculated in S300 is returned and the determination ends.

  In S325, (x ′, y ′) calculated in S310 is returned, and the NLOS determination is terminated. Returning (x ′, y ′) means that a set of measurement values (TOA measurement value, AOA measurement value) that was not used in the calculation of (x ′, y ′) is the position of the target node 2. It has been removed from the target used for estimation.

  Returning to FIG. 2, in S <b> 120, the returned value is set as the estimated position of the target node 2. With the above, the NLOS discrimination and the estimated position calculation are finished.

  Next, a computer simulation is performed using the NLOS discrimination method (FIG. 2) executed by the central processing unit 4 in the present embodiment, and the result of comparing the characteristics will be described. FIG. 5 shows a diagram of the sensor field 1 which is a region where position estimation is performed in the computer simulation. FIG. 5 shows a two-dimensional space having a vertical length and a horizontal length of 10.0 m, respectively, and nine sensor nodes 3-i (N = 1,..., 9), of which the sensor node 3-under the NLOS environment 1, 3-6, 3-7. Further, although not shown, the target nodes 2 are arranged on the grid points in increments of 0.1 [m] in the x direction and the y direction in the entire sensor field 1 in FIG. Assume that position estimation using the NLOS discrimination of the embodiment is performed.

Here, the measured value at each sensor node 3-i is an averaged value after measuring for the determined number of measurements. The measured value after averaging when the number of pre-measurements is M and the measurement value of the j-th pre-measurement number at the i-th sensor node 3-i is d _{i, j} is expressed by the following equation. .

By performing this operation, it is possible to suppress the variance of the noise included in the measurement value, and the estimation accuracy is improved each time the number of measurements M is increased. In the computer simulation, the number of prior measurements M = 30.

  The frequency used for position estimation was set to 500 [MHz], which is Ultra Wide Band (UWB). UWB has high temporal resolution and is suitable for position estimation using TOA that requires high temporal resolution (however, the application target of the present invention is not limited to UWB). Also, since the measurement error is relatively small, it is often used for position estimation. As the line-of-sight (LOS) error of the TOA measurement value in the indoor environment, a value obtained from the actual measurement value shown in Non-Patent Document 2 is used. According to Non-Patent Document 2, the LOS error of TOA follows a Gaussian distribution, and its mean and variance are expressed by the following equations.

Here, d _i is the distance between the sensor node 3-i and the target node 2, and m _{TOA, LOS} = 0.21 m and σ _{TOA, LOS} = 0.269 m when the operating frequency is 500 [MHz].

  AOA LOS error generally follows a Gaussian distribution with a mean of zero. In the computer simulation, the variance of the LOA error of AOA was set as follows:

Further, the weights γ _{i, TOA} , γ _{i, and AOA} of the LS method in the HTA method of equation (3) are the reciprocals of equations (9) and (10), respectively.

Secondly, non-line-of-sight (NLOS) errors in TOA and AOA measurements are often given in a single scattering model. The single scattering model is a model in which the radio wave transmitted from the target node 2 reaches the sensor node 3-i after being scattered only once in the scatterer. In an actual environment, the radio wave may be received by the sensor node 3-i after being scattered a plurality of times, but the signal power is greatly attenuated each time the scattering occurs. The signal that arrives cannot be received due to power reduction. The distance between the sensor node 3-i and the target node 2 is D, the position of the sensor node 3-i is (-D / 2, 0), the actual position of the target node 2 is (-D / 2, 0), and the scatterer Where (p, q) is the NLOS error m _{TOA, NLOS of TOA} and NLOS error m _{AOA, NLOS} of _AOA are expressed by the following equations.

From the equations (11) and (12), the NLOS error of TOA and AOA depends on the position (p, q) of the scatterer. Various models exist such as CSM (Circular Scattering Model) that assumes an outdoor environment and ESM (Elliptical Scattering Model) that assumes an indoor environment in order to determine the position (p, q) of the scattering object. In computer simulation, GSDM (Gaussian Scatter Density Model) shown in Non-Patent Document 3 is used. This model assumes that the distribution of scattered objects follows a two-dimensional Gaussian distribution centered on the target node 2, and is a model that has been measured in both indoor and outdoor environments. In the computer simulation, the variance σ _s of the two-dimensional Gaussian distribution in GSDM is expressed as follows.

Next, a minute value δ used in the IMR method which is an NLOS discrimination method is set. In the NLOS discrimination of this embodiment, since the IMR method is performed in two cases, when only the TOA measurement value is used (S105) and when both the TOA measurement value and the AOA measurement value are used (S115), the minute value δ is There are two types. Minute value δ when using only TOA measurement value δ _TOA (used as δ in S215 of FIG. 3) and minute value δ _HTA when using both TOA measurement value and AOA measurement value (As δ in S315 of FIG. 4) Used) was defined as follows:

In equation (14), ε _i is a normalized residual ε ′ when i TON measurement values are used in S210 when i <N, and a normality calculated in S200 when i = N. The residual η. Therefore, _δTOA used in the case of N _d <N−1 is η−ε ′ calculated in the previous S215.
Is equivalent to 1/5 of the absolute value of. That is, when N _d <N−1, it is determined whether or not the amount of change in the normalized residual in S215 this time is smaller than 1/5 of the absolute value of the amount of change in the normalized residual in S215 last time. Will be. Since the TOA measurement value excluded in S218 is a measurement value with a large error, in most cases, the residual ε ′ decreases with each execution of S210. Therefore, the larger the reduction amount of the normalization error, the larger the value of _δTOA . Will be big.

  A root mean square error (RMSE) is used for performance evaluation of the position estimation system. RMSE is a standard deviation between the true position of the target node 2 and the estimated position. The RMSE when the position is estimated L times is given by the following equation.

Here, (X, Y) is the true position of the target node 2, and (x _{R, l} , y _{R, l} ) is the estimated position obtained by the i-th position estimation. In the computer simulation, the estimated number of times L = 1000.

  FIG. 6 shows the RMSE in the entire sensor field 1 of FIG. 5 calculated by computer simulation. Here, as an existing method, FIG. 7 shows, as a comparative example, RMSE when the IMR method is applied to the TOA method for the sensor field 1 under the same conditions as in FIG.

  FIG. 8 shows a comparative example of RMSE when the IMR method is simply applied to the HTA method for the sensor field 1 under the same conditions as in FIG. Simply applying the IMR method to the HTA method does not always execute the IMR method using only the TOA measurement value, but always executes the IMR method using both the TOA measurement value and the AOA measurement value. That is. More specifically, for the processing of S200 and S210 in FIG. 3, not only the TOA measurement value but also all the AOA measurement values that are paired with the TOA measurement value to be used are used, and the equations (3) to (6) It is sufficient to make a change such that the estimated position of the target node 2 and the normalized residual are calculated by using.

  In FIGS. 6 to 8, the portion with a low RMSE is shown in white and the portion with a high RMSE is shown in black. Since the position estimation accuracy is better as the RMSE is lower, it can be said that the estimation accuracy is better as the number of white portions is increased. Comparing FIG. 6, FIG. 7 and FIG. 8, it can be seen that there are many parts shown in white in FIG.

  In FIG. 8, a range near the center of the sensor field (more specifically, a range of 5.0 m ≦ X ≦ 8.0 m and 2.0 m ≦ Y ≦ 8.0 m, where X = 5.0 m, In a range excluding a circle with a radius of 1.0 m centered at Y = 5.0 m), the RMSE is significantly larger than that in FIG. This is because when the IMR method is simply applied to the HTA method as described above, a plurality of NLOS measurement values (TOA measurement values and AOA measurement values of the sensor node under the NLOS environment) cannot be removed well, and an error occurs. This is because the estimated position is calculated while leaving a plurality of NLOS measurement values having large values.

  In fact, in the example of FIG. 8, the measured values (TOA measured values, 3A, Among the AOA measurement value sets), only two or less sets can be removed, and any one of the sensor nodes 3-2 to 3-5, 3-8, and 3-9 that are not in the NLOS environment. One or more measurement values (a set of TOA measurement values and AOA measurement values) have been removed.

  This is because in the position estimation using the AOA, the estimated position varies greatly due to a large error due to the NLOS environment. As described above, in the IMR method, when the sensor node in the NLOS environment is determined by comparing the residual obtained from the estimated position using the AOA, the accuracy of the determination is deteriorated if there are two or more sensor nodes in the NLOS environment. To do.

  On the other hand, in FIG. 6 which is a result when the method of the present embodiment is used, it can be seen that the RMSE is remarkably lowered in the range near the center of the sensor field as compared with the result of FIG. . This is because a plurality of NLOS measurement values are removed by NLOS discrimination using only the TOA measurement values (S105, S110), and the NLOS measurement values remaining in the NLOS discrimination (S115) in the HTA method to be subsequently performed (S115). (TOA measurement value, AOA measurement value) becomes one set or zero set, and as a result, NLOS discrimination is correctly performed.

  In other words, the central processing unit 4 is executed at almost all positions within the range near the center of the sensor field described above, by repeating the loop of S210, S215, and S220 three or more times in the process of FIG. The TOA measurement value of the sensor node 3-i (including a plurality of sensor nodes in the NLOS environment) is removed from the object used for position estimation of the target node 2, and then finally removed in S110 of FIG. 2 via S225. Return the measured TOA value to the target. As a result, the TOA measurement values of two or more sensor nodes 3-i (including a plurality of sensor nodes in the NLOS environment) are removed in S105 and S110. Further, the central processing unit 4 has the NLOS measurement values (TOA measurement values, TOA measurement values) remaining in the target used for position estimation of the target node 2 in the processing of FIG. If the AOA measurement value) is only one set, the estimated position calculated by removing the one set of NLOS measurement values in S325 is returned, and if the NLOS measurement value remaining in the target is zero set, S320 Returns the estimated position calculated using all pairs remaining in.

  6 to 8, the RMSE averaged over the entire sensor field is 0.23985 m in FIG. 6, 0.67696 m in FIG. 7, and 0.33071 m in FIG. 8. From this, it can be seen that, in the entire sensor field, the present embodiment improves the position estimation accuracy over the existing method. In addition, in FIG. 6 as a whole, the number of sensor nodes in the NLOS environment from which the TOA measurement value was removed in S105 of FIG. 2 (the number of removals after the return in S110) was 2.0008 on average. It was. This number is smaller than the number of sensor nodes under the NLOS environment, and the sensor nodes under the NLOS environment cannot be completely removed. However, even if it is an NLOS measurement value, the included error may be very small. In this case, the position estimation accuracy is improved by using the measurement value without removing the measurement value. For this reason, the NLOS measurement value that has not been removed in S105 has a small error, and does not deteriorate the abnormal value determination accuracy in S115 and the position estimation accuracy in S120.

  In this embodiment, the central processing unit 4 functions as an example of the first removal unit by executing S105 and S110, functions as an example of the second removal unit by executing S115, and executes S120. This functions as an example of an estimated position acquisition unit.

  In the present embodiment, as described above, the TOA measurement values of a plurality of sensor nodes in the NLOS environment based on only the TOA measurement value among the TOA measurement value and the AOA measurement value in S105 and S110 in advance. However, there is a high possibility that the target node 2 is excluded from the target used for position estimation of the target node 2. Therefore, at the start of execution of S115 using both the TOA measurement value and the AOA measurement value, the number of TOA measurement values and AOA measurement values of the sensor node in the NLOS environment is reduced from the initial value. Accuracy is improved.

In the present embodiment, in the NLOS determination using the IMR method in the HTA method, the NLOS determination is first performed using the measured value of only the TOA, and after removing a plurality of sensor nodes in the NLOS environment, the TOA is continuously performed. And NLOS discrimination using the measured value of AOA. In such a discriminating method, the sensor nodes of a plurality of NLOS environments are removed by NLOS discrimination using only the measured value of TOA. Therefore, in the subsequent NLOS discrimination using both TOA and AOA, it is used for discrimination. The measurement value included in the measurement value is often one or less in the NLOS environment, and it is possible to reduce errors in NLOS determination.
(Other embodiments)
In the above embodiment, the central processing unit 4 (corresponding to an example of a processing unit) is realized as a separate device different from the target node 2 and the sensor node 3-i. However, this need not be the case. For example, the central processing unit 4 may not be prepared, and the sensor node 3-i and the target node 2 may have the function of the central processing unit 4.

  In the above embodiment, position estimation is performed in a two-dimensional space, but position estimation may be performed in a three-dimensional space.

In the above-described embodiment, each sensor node 3-i calculates the distance d _i between the target node 2 and itself based on the TOA of the received signal, and transmits the calculated distance d _i to the central processing unit 4. It is supposed to be. However, the TOA itself of the received signal may be transmitted to the central processing unit 4. In that case, the central processing unit 4 may calculate the distance d _i based on the received TOA.

In the above embodiment, each sensor node 3-i measures the distance d _i between the target node 2 and its own device using the TOA method. However, not the case, for example, using other distance measurement methods such as RSS method or TDOA method may be adapted to perform the measurement and transmission of the distance d _i between the target node 2 own device. In that case, each sensor node 3-i may transmit the RSS itself of the received signal or the TDOA itself to the central processing unit 4 instead of the distance d _i . In that case, in that case, the central processing unit 4 may calculate the distance d _i based on the received RSS or TDOA.

  That is, each sensor node 3-i transmits a set of distance information that can calculate the distance from the target node (2) to its own device and arrival direction information that can calculate the arrival direction of the signal. I just need it.

  In the above embodiment, in the second IMR method, NLOS discrimination using both TOA and AOA is performed only once, but this may be performed a plurality of times.

  In the above embodiment, the least square method is used to calculate the estimated position of the target node 2, but other calculation methods may be used.

  INDUSTRIAL APPLICABILITY The present invention can be used as a method for appropriately removing an abnormal value and improving position estimation accuracy in position estimation for performing distance measurement and angle measurement in a sensor network.

DESCRIPTION OF SYMBOLS 1 ... Sensor field 2 ... Target node 3-i ... Sensor node 4 ... Central processing unit

Claims (2)

Receives a signal transmitted from the target node (2), and based on the received signal, distance information (d _i ) capable of calculating a distance from the target node (2) to the own device, and arrival of the signal A plurality of sensor nodes (3-i) that calculate a pair with direction-of-arrival information (θ _i ) capable of calculating a direction;
Processing means (4),
The processing means (4) is a measurement value acquisition means (S100) for acquiring a set of distance information (d _i ) and arrival direction information (θ _i ) calculated at each of the plurality of sensor nodes (3-i). When,
Of all the sets of distance information (d _i ) and arrival direction information (θ _i ) acquired by the acquisition unit (S100), only the distance information (d _i ) is used for position estimation of the target node (2). A first removal means (S105, S110) for selecting distance information (d _i ) having a large error using the target and including the selected distance information (d _i ) from the target;
After the distance information (d _i ) having a large error is removed from the target by the first removing means (S105, S110), it is combined with any one of a plurality of distance information (d _i ) remaining on the target. A plurality of direction-of-arrival information (θ _i ) is added to the target, and using the target after the addition, distance information (d _i ) and direction-of-arrival information (θ _i ) having a large error are added A second removing means (S115) for removing the set;
The estimated position of the target node calculated using the target after the pair of the distance information (d _i ) and the arrival direction information (θ _i ) having a large error is removed by the second removing unit (S115) is acquired. An estimated position acquisition unit (S120). The first removal means (S105, S110)
When the number of the distance information (d _i) included in the subject is a +1 N _d, estimated position of the target node (2) using N _d +1 distance information included in the object (d _i) and calculating the residual eta, further, among the _N d +1 pieces of distance information included in the object _{(d i),} using each of the _{N d} number of measurements of the _N d +1 _Street, N d +1 calculating the target node estimated position and N residuals _d +1 street (2), as the (epsilon), the residual of the N _d +1 ways residual becomes minimum value among (epsilon) Ipushiron' and A comparison with the residual η
The above process is repeated while removing distance information (d _i ) that has not been used for the calculation of the residual ε ′ from the target until η−ε ′ becomes smaller than a set value (δ _TOA ),
The position estimation system according to claim 1, wherein when η−ε ′ is smaller than a set value (δ _TOA ), the distance information (d _i ) last removed from the object is returned to the object.