JP2024007217A - Positioning device, leaning device, positioning method, learning method, and positioning program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain stable positioning results.

SOLUTION: A positioning device (1) is provided, comprising an area estimation unit (103) configured to estimate an area with a measurement point used to measure positioning data within a building using the positioning data, and a position estimation unit (105) configured to estimate the position of the measurement point in the estimated area.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2024,JPO&INPIT

Description

The present invention relates to a positioning device and the like that perform indoor positioning.

Various studies have been carried out regarding indoor positioning technology. For example, Non-Patent Document 1 listed below discloses a technique for estimating on which floor in a building having an atrium a BLE (Bluetooth Low Energy, Bluetooth is a registered trademark) signal is measured. The atrium is a large indoor courtyard-like space with a high ceiling.

Hongxia Qi etc., "BLE-based floor positioning method for multi-level atrium spatial environments", Acta Geodaetica et Geophysica, 2021, 56:p.471-p.499

The technique disclosed in Non-Patent Document 1 has the problem that it is necessary to provide a large number of wireless access points within a building, which increases the installation cost. Furthermore, it is assumed that positioning accuracy may decrease depending on the structure of the building, the arrangement of wireless access points, etc. In addition, other indoor positioning techniques are known, such as those that use magnetism and methods such as dead reckoning, but these methods may reduce positioning accuracy when the space to be positioned is large. . In other words, these methods have the problem that the positioning results are not stable depending on whether the space of the positioning target is wide or narrow. One aspect of the present invention aims to realize a positioning device and the like that can obtain stable positioning results.

In order to solve the above problems, a positioning device according to one aspect of the present invention provides an area estimation method that uses the positioning data to estimate an area where a measurement point where positioning data is measured in a building exists. and a position estimation unit that estimates the position of the measurement point in the estimated area.

In order to solve the above problems, a learning device according to one aspect of the present invention uses the difference between the atmospheric pressure measured at each point in a building and the atmospheric pressure measured at a predetermined reference point to and a teacher data acquisition unit that acquires teacher data in which whether or not the point is an atrium area is associated as a correct answer label, and learning using the teacher data to calculate the difference in air pressure between the measurement point and the reference point. A learning unit that generates an estimation model for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area.

In order to solve the above problems, a positioning method according to one aspect of the present invention is a positioning method executed by one or more information processing devices, in which a measurement point where positioning data is measured in a building is an area estimation step of estimating an existing area using the positioning data; a position estimation step of estimating the position of the measurement point in the estimated area based on magnetism measured at the measurement point; ,including.

In order to solve the above problems, a learning method according to one aspect of the present invention is a learning method that is executed by one or more information processing devices, in which the atmospheric pressure measured at each point in a building and a predetermined a teacher data acquisition step of acquiring teacher data in which the hierarchy of each point and whether or not the point is an atrium area are associated as correct answer labels with the difference between the atmospheric pressure measured at the reference point; a learning step of generating an estimation model for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area from the pressure difference between the measurement point and the reference point by learning using the data; include.

According to one aspect of the present invention, it is possible to obtain stable positioning results.

FIG. 1 is a block diagram showing an example of the configuration of main parts of a positioning device according to an embodiment of the present invention. FIG. 1 is a block diagram showing an example of the configuration of main parts of a learning device according to an embodiment of the present invention. FIG. 1 is a diagram showing an overview of a positioning system including the positioning device and the learning device. This is a histogram of air pressure differences in each of a plurality of rooms located on the same floor. It is a flowchart which shows an example of the process which the said positioning device performs. It is a flowchart which shows an example of the process which the said learning device performs.

〔System configuration〕
The positioning system 5 according to this embodiment will be explained based on FIG. 3. FIG. 3 is a diagram showing an overview of the positioning system 5. As shown in FIG. As illustrated, the positioning system 5 includes a positioning device 1, a learning device 2, and a measuring device 3. Further, FIG. 3 also shows the configuration of the building A that is the object of positioning by the positioning system 5.

Building A is a multi-story building including an atrium area. Specifically, building A is a three-story building with first to third floors, and rooms A1 to A6 are provided within building A. Among these, room A1 is an open-air room spanning three floors, the first to third floors. Among the other rooms A2 to A6, A2 and A3 are located on the first floor, A4 on the second floor, and A5 and A6 on the third floor. Further, the other rooms A2 to A6 include those that are adjacent to the room A1 (A2, A3, A5, A6) and those that are not adjacent to the room A1 (A4). Further, the upper part of the room A3 is a part of the room A1, so that the second floor part of the room A1 is wider than the first and third floor parts. Further, although not shown in the drawing, a staircase is provided in the room A1 so that the user can move from the first floor to the third floor within the room A1.

Person B is inside building A (more specifically, inside room A1). As shown in the figure, person B has a measurement device 3 that measures positioning data. The positioning device 1 acquires the positioning data measured by the measuring device 3, and uses the data to measure the position of the person B in the building A, that is, the measurement point of the positioning data. Although details will be described later, multiple types of data are used for positioning. Each type of data may be measured using different measuring devices. Further, the measuring device 3 may have a measurement as its main function, or may have another function as its main function. For example, a smartphone or the like equipped with various sensors can also be used as the measuring device 3.

Although details will be described later, when performing the above positioning, the positioning device 1 first estimates the area where the person B is present, and then estimates the position of the person B in the estimated area. For example, in the example of FIG. 3, the positioning device 1 estimates that the area where the person B is located is the first floor in the room A, and then positions the person B in the first floor in the room A. Estimate.

This makes it possible to obtain stable positioning results even when using a positioning method, such as a positioning method using magnetism, in which positioning accuracy decreases when the positioning target space is wide. This area estimation uses an estimation model that has been machine learned in advance, and this estimation model is generated by the learning device 2.

The positioning system 5 can also be used for safety management of workers in various facilities such as waste incineration facilities and various factories. For example, in waste incineration facilities, the furnace room where the incinerator is located is an atrium, and around it there are multiple smaller rooms than the furnace room, such as a ventilation room and a high-voltage electricity room. , has a similar configuration to Building A. The furnace room is larger than a typical living space, and inside is a grid-like metal structure called grating, which provides installation locations for various equipment and passages for maintenance of the equipment. There is. Workers sometimes enter the furnace room for maintenance, etc., but if a worker has an accident or gets injured and becomes unable to move inside the vast furnace room with poor visibility, it will take time to find the problem. It is assumed that this will require a In this regard, if the worker is equipped with the measuring device 3, the positioning system 5 can estimate the worker's position and detect the worker who is stuck, or quickly rescue the detected worker. becomes possible.

[Configuration of positioning device]
The configuration of the positioning device 1 will be explained based on FIG. 1. FIG. 1 is a block diagram showing an example of the configuration of main parts of the positioning device 1. As shown in FIG. As illustrated, the positioning device 1 includes a control unit 10 that centrally controls each part of the positioning device 1, and a storage unit 11 that stores various data used by the positioning device 1. The positioning device 1 also includes a communication unit 12 for the positioning device 1 to communicate with other devices, an input unit 13 for receiving input of various data to the positioning device 1, and an output for the positioning device 1 to output various data. 14. The control unit 10 also includes a data acquisition unit 101, a preprocessing unit 102, an area estimation unit 103, an area integration unit 104, and a position estimation unit 105. The storage unit 11 stores positioning data 111, an estimation model 112, and verification magnetic data 113.

The data acquisition unit 101 acquires positioning data 111. The acquired positioning data 111 is stored in the storage unit 11. The positioning data 111 includes various data used for area estimation by the area estimation unit 103 and various data used for positioning by the position estimation unit 105. Details of these data will be described later.

The preprocessing unit 102 performs preprocessing on the positioning data 111 to make it data that can be input to the estimation model 112 . Details of the preprocessing will be described later. Note that the data acquisition unit 101 may acquire preprocessed data, and in this case, the preprocessing unit 102 is omitted.

The area estimating unit 103 uses the positioning data 111 to estimate an area in which a measurement point where the positioning data 111 was measured exists in a building that is a positioning target. More specifically, the area estimating unit 103 estimates the area where the measurement point is located based on the output value obtained by inputting the positioning data 111 preprocessed by the preprocessing unit 102 into the estimation model 112. . Details of the estimation model 112 will be described later.

Furthermore, the area estimating unit 103 may estimate the probability that the measurement point exists in each of a plurality of areas where the measurement point may exist. For example, the area estimation unit 103 can perform the estimation by using the estimation model 112 that outputs the probability.

The area integrating unit 104 integrates a predetermined number of areas with higher probabilities estimated by the area estimating unit 103 to form an integrated area. For example, in building A in Figure 3, the probability that the measurement point is on the first floor of room A1 is 80%, the probability that the measurement point is outside room A1 on the first floor (that is, inside room A2) is 15%, and the probability that the measurement point is on the second floor is on the first floor. Suppose that the probability of being outside room A1 (that is, inside room A4) is estimated to be 5%. In this case, if the predetermined number is 2, the area integration unit 104 sets the area where the first floor portion of the room A1 and the room A2 are integrated, that is, the entire first floor, as the integrated area.

The position estimating unit 105 estimates the position of the measurement point in the area estimated by the area estimating unit 103. Furthermore, when the area integration unit 104 has generated an integrated area, the position estimation unit 105 estimates the position of the measurement point in the integrated area. Although details will be described later, for example, the position estimating unit 105 may perform the above estimation using the verification magnetic data 113.

As described above, the positioning device 1 uses the area estimating unit 103 and the area estimating unit 103 to estimate the area where the measurement point where the positioning data 111 was measured in the building exists, using the positioning data 111. and a position estimating unit 105 that estimates the position of the measurement point in the estimated area.

According to the above configuration, the area in which the measurement point exists in the building is estimated, and then the position of the measurement point in the estimated area is estimated. As a result, positioning can be performed after narrowing down the positioning target area, so positioning can be performed with stable accuracy.

Further, according to the above configuration, there is a low possibility that a position far away from the actual measurement point will be estimated. Therefore, the positioning device 1 can be suitably applied to positioning workers in various facilities. For example, suppose that an unexpected accident or the like occurs in a facility and it becomes necessary to locate and rescue a worker, and the position of the worker is determined using the positioning device 1. In this case, even if the accuracy of position estimation based on magnetism is not sufficient, if the area is estimated correctly, the area corresponding to the estimated position will be searched and the worker will be immediately rescued. be able to.

Further, as described above, the area estimation unit 103 may estimate the probability that the measurement point exists in each of a plurality of areas where the measurement point may exist. In this case, the position estimation unit 105 may estimate the position of the measurement point in an integrated area that integrates a predetermined number of areas with higher estimated probabilities.

In the positioning device 1, if the result of area estimation performed before position estimation is incorrect, there is a possibility that the positioning result will indicate a position far away from the actual measurement point. In this regard, according to the above configuration, since positioning is performed within an integrated area that integrates a predetermined number of areas with high probabilities, there is a possibility that the positioning result will indicate a position far away from the actual measurement point. can be reduced.

[Configuration of learning device]
The configuration of the learning device 2 will be explained based on FIG. 2. FIG. 2 is a block diagram showing an example of the main part configuration of the learning device 2. As shown in FIG. As illustrated, the learning device 2 includes a control section 20 that centrally controls each section of the learning device 2, and a storage section 21 that stores various data used by the learning device 2. The learning device 2 also includes a communication unit 22 for the learning device 2 to communicate with other devices, an input unit 23 for receiving input of various data to the learning device 2, and an output for the learning device 2 to output various data. 24. The control unit 20 also includes a data acquisition unit 201, a teacher data generation unit 202, a teacher data acquisition unit 203, and a learning unit 204. The storage unit 21 stores learning data 211, teacher data 212, and estimation model 112.

The data acquisition unit 201 acquires various data used for learning by the learning unit 204. The acquired data is stored in the storage unit 21 as learning data 211. The learning data 211 is the source data of the teacher data 212.

The teacher data generation unit 202 generates teacher data 212 using the learning data 211. The generated teacher data 212 is stored in the storage unit 21. Note that a specific method for generating the teacher data 212 will be described later.

The teacher data acquisition unit 203 acquires the teacher data 212 that the learning unit 204 uses for learning. The learning device 2 includes the teacher data generation section 202 as described above, and since the teacher data 212 generated by the teacher data generation section 202 is stored in the storage section 21, the teacher data acquisition section 203 acquires data from the storage section 21. What is necessary is to acquire the teacher data 212. If the learning device 2 does not include the teacher data generation unit 202, the teacher data acquisition unit 203 may acquire the teacher data 212 generated by another device from the other device, or the learning device 2 may acquire the teacher data 212 generated by another device. The teacher data 212 inputted by the user via the input unit 23 or the like may be obtained.

The learning unit 204 generates the estimated model 112 by learning using the teacher data 212. Note that the details of learning will be described later. The generated estimation model 112 is stored in the storage unit 21. Further, the generated estimation model 112 is provided to the positioning device 1. The estimated model 112 may be provided to the positioning device 1, for example, by communication via the communication unit 22, or may be provided manually by the user.

[Area estimation based on pressure difference]
Since atmospheric pressure has the characteristic of varying depending on height, measured atmospheric pressure can be used to estimate the floor. However, atmospheric pressure fluctuates depending on various factors such as weather. Therefore, in area estimation, it is preferable to use the difference between the measured atmospheric pressure and the atmospheric pressure measured at a predetermined reference point, that is, the atmospheric pressure difference, instead of using the measured atmospheric pressure as is. Specifically, the atmospheric pressure difference may be included as one of the explanatory variables of the estimation model 112.

Additionally, as a result of experiments conducted by the inventors of the present application, it has been found that it is possible to identify multiple rooms located on the same floor based on air pressure differences. This will be explained based on FIG. 4. FIG. 4 is a histogram of air pressure differences in each of a plurality of rooms located on the same floor. More specifically, FIG. 4 shows histograms D1 and D2 showing the frequency distribution of the pressure difference calculated from the pressure measured while moving along different routes within a certain floor of the furnace room in the waste incineration facility. It shows. Further, FIG. 4 shows a histogram D3 showing the frequency distribution of the atmospheric pressure difference calculated from the atmospheric pressure measured while moving in one room other than the furnace room in the above-mentioned hierarchy.

Since the distributions of histograms D1 and D2 generally overlap, it can be seen that the pressure difference has a similar distribution regardless of the movement path within the furnace chamber. On the other hand, there is a deviation between the distribution shown in histogram D3 and the distributions shown in histograms D1 and D2.

Since the histograms D1 to D3 all show the distribution of atmospheric pressure differences calculated from the atmospheric pressures measured on the same floor, it is considered that the above-mentioned deviation is caused by factors other than the floor. In other words, the histograms D1 to D3 in Fig. 4 indicate that there is a possibility that the furnace room (area with an atrium) located on the same floor can be distinguished from a room other than the furnace room (an area that is not an atrium) based on the pressure difference. This shows that it is possible to identify multiple rooms located on the same floor based on the difference.

As described above, the building to be positioned may be a multi-story building including an atrium area and a non-atrium area. In this case, the positioning data 111 may include data indicating the atmospheric pressure measured at the measurement point. Then, based on the difference between the atmospheric pressure measured at the measurement point and the atmospheric pressure measured at a predetermined reference point, the area estimation unit 103 determines the hierarchy of the measurement point and whether or not the measurement point is in an atrium area. may be estimated.

According to the above configuration, it can be expected that a valid estimation result can be obtained regarding the hierarchy of the measurement point and whether or not the measurement point is in an atrium area. In addition, according to the above configuration, the estimation is based on the difference between the measured atmospheric pressure and the atmospheric pressure measured at a predetermined reference point, so it is stable regardless of changes in atmospheric pressure due to weather etc. area estimation.

In addition, the teacher data 212 acquired by the teacher data acquisition unit 203 includes the difference between the atmospheric pressure measured at each point in the building and the atmospheric pressure measured at a predetermined reference point, the floor level of each point, and the atrium at the point. The correct answer label may be associated with whether or not the area is in the area.

In this case, the learning unit 204 performs learning using the teacher data 212 to estimate the hierarchy of the measurement point and whether or not the measurement point is in an atrium area based on the pressure difference between the measurement point and the reference point. Estimated model 112 can be generated.

[Area estimation based on received signal strength indication (RSSI)]
For indoor positioning, wireless equipment such as Wi-Fi (registered trademark) and BLE access points may be used. Such a positioning method takes advantage of the fact that in wireless communication using wireless equipment, as the distance between the base station and the terminal increases, the strength of the received signal at the terminal attenuates. In this positioning method, positioning is performed by using the strength of a received signal from a specific access point, measured by a terminal, as a feature indicating the location where the terminal is located. The received signal strength of Wi-Fi or BLE can also be used for area estimation by the positioning device 1.

For example, when performing area estimation in a building with a vast indoor space such as a waste incineration facility, the positioning device 1 uses the received signal strength of LPWA (Low Power Wide Area-network) to estimate the area. An estimate may be made. Specifically, the received signal strength of the LPWA may be included as one of the explanatory variables of the estimation model 112.

LPWA is a wireless communication standard used for longer distance communication than Wi-Fi or BLE. While Wi-Fi and BLE use the 5GHz and 2.4GHz frequency bands, LPWA uses the lower 920MHz band. In general, the lower the frequency, the more suitable it is for long-distance communication, so LPWA is suitable for area estimation inside a building with a large indoor space. Furthermore, when applying LPWA, there is an advantage that the number of base stations to be placed within a building can be reduced compared to when applying Wi-Fi or BLE.

In the past, LPWA was mainly considered for use in outdoor positioning, and was used in buildings with large indoor spaces, especially in areas where a large number of equipment is installed, such as in waste incineration facilities. It has not been verified whether LPWA can be used for estimation. However, experiments conducted by the inventors of the present invention have confirmed that the received signal strength of LPWA can be used for area estimation in waste incineration facilities.

As described above, the positioning data 111 may include the LPWA received signal strength measured at the measurement point. In this case, the area estimation unit 103 estimates the area where the measurement point is located based on the received signal strength of the LPWA. LPWA is a standard that assumes long-distance communication compared to Wi-Fi, BLE, etc., so the above configuration that estimates the area based on the received signal strength of LPWA has a large internal space. Suitable for positioning inside buildings.

[About the estimation model]
The estimation model 112 may be any model that can estimate an area according to the positioning data 111. In other words, the estimation model 112 may be a model that uses the positioning data 111 or a feature amount generated using the same as an explanatory variable, and uses the area where the measurement point of the positioning data 111 exists as an objective variable.

As explanatory variables, for example, the above-mentioned atmospheric pressure difference, received signal strength in any communication technology such as Wi-Fi, BLE, or LPWA, etc. may be applied. For example, by using both the pressure difference and the received signal strength of the LPWA as explanatory variables, it is possible to estimate the area with high accuracy within a building such as an incineration facility that has a vast internal space. Of course, the explanatory variables may be those that are related to the area, and are not limited to the above example. Furthermore, the number of explanatory variables to be used is not particularly limited.

Furthermore, although the estimation algorithm used in the estimation model 112 is not particularly limited, it is preferable that the estimation model 112 is a machine learning model that can be expected to have high estimation accuracy. The estimation model 112 may be a linear model or a nonlinear model. For example, a decision tree-based estimation algorithm such as random forest may be applied, an estimation algorithm using a neural network such as a multilayer perceptron may be applied, or a regression-based estimation algorithm such as logistic regression may be applied. You may.

〔Preprocessing〕
The positioning data 111 acquired by the data acquisition unit 101 may include data that cannot be directly input into the estimation model 112. The preprocessing unit 102 preprocesses such data to make it data that can be input to the estimation model 112. For example, the preprocessing unit 102 may perform interpolation processing on the positioning data 111.

The interpolation process is performed to match the number of data of a plurality of explanatory variables in the estimation model 112. For example, assume that the explanatory variables of the estimation model 112 include the atmospheric pressure difference and the received signal strength of the LPWA. Since the measurement of atmospheric pressure and the measurement of received signal strength are performed by different sensors, in this case, one measurement value may exist at a certain time, but the other measurement value may not exist. If one of the measured values does not exist, the input data to the estimation model 112 at that time will be incomplete. Therefore, the preprocessing unit 102 interpolates the other measured value to obtain data that can be input to the estimation model 112. For example, when a measured value at a certain time is a missing value, the preprocessing unit 102 may interpolate the missing value with the most recent measured value at that time.

Further, at points where radio waves from the base station cannot be received at all, the received signal strength cannot be measured, resulting in missing measured values. For such points, the preprocessing unit 102 may interpolate the missing values using a sufficiently small received signal strength. For example, the preprocessing unit 102 may interpolate the missing value with a value of −100 dbm for a portion of the received signal strength value included in the positioning data 111 that is a missing value due to unmeasurability.

Moreover, a certain absolute error exists in each individual barometric pressure sensor used to measure atmospheric pressure. Therefore, when the positioning data 111 includes an atmospheric pressure value, the preprocessing unit 102 may perform a process of subtracting an absolute error from the value as an offset. The same applies to the atmospheric pressure measured at a predetermined reference position. The preprocessing unit 102 then calculates the difference between the atmospheric pressure measured at the measurement point and the atmospheric pressure measured at a predetermined reference position.

Further, the preprocessing unit 102 may perform scaling on the positioning data 111 to generate data to be input to the estimation model 112. For example, the preprocessing unit 102 may perform minimum-maximum scaling on the positioning data 111, or may normalize the positioning data 111. Note that the data to be scaled and normalized may be data after the above interpolation.

Similarly, the learning data 211 acquired by the data acquisition unit 201 may include data that cannot be directly input to the estimation model 112 and cannot be used as the teacher data 212 as is. In this case, the teacher data generation unit 202 generates data that can be input to the estimation model 112 through processing similar to that of the preprocessing unit 102.

Further, the teacher data generation unit 202 may perform a process of increasing the number of received signal strength data. For example, the teacher data generation unit 202 may generate a new received signal strength by adding a noise component to the received signal strength included in the learning data 211. The noise component may be, for example, Gaussian noise generated based on the average value and standard deviation of each received signal strength included in the learning data 211.

In this way, when the training data 212 includes the received signal strength of the LPWA measured at each point in the building, the training data generation unit 202 uses data obtained by adding a noise component to the received signal strength of the LPWA. The teacher data 212 may also be generated by Regarding the received signal strength of LPWA, there is a possibility that the universality of the teaching data 212 is insufficient during learning, but according to the above configuration, the universality can be compensated for by increasing the variations of the teaching data 212.

Furthermore, since the received signal strength changes over time, it is ideal to perform long-term measurements at all measurement points. In this regard, according to the above configuration that adds a noise component to the measured received signal strength, it is possible to simulate the temporal fluctuations of the received signal strength, so the received It is possible to generate training data 212 that takes into account temporal fluctuations in signal strength.

[About magnetic positioning]
Indoors, magnetic characteristics are created at each location by retaining residual magnetism in iron structures. The position estimation unit 105 may perform positioning using this feature. In this case, the data acquisition unit 101 acquires magnetic data indicating magnetism measured at the measurement point as positioning data 111. For example, the data acquisition unit 101 may acquire data indicating the magnetic strength in three axial directions orthogonal to each other. In addition, magnetism is measured in advance at each point in the building to be positioned, and the data indicating the magnetic strength measured at each point is associated with the measurement point and stored in the storage unit 11 as verification magnetic data 113. I'll keep it. The magnetic characteristics shown in the verification magnetic data 113 can also be expressed as a magnetic fingerprint.

Thereby, the position estimating unit 105 can estimate the measurement point by comparing the magnetic data included in the positioning data 111 and the verification magnetic data 113. Note that the wider the area to be estimated, the higher the possibility that magnetic fingerprints will match at multiple points, and the higher the possibility that erroneous estimation results will be output. Therefore, in order to improve the estimation accuracy by the position estimating unit 105, it is important to narrow down the area to be estimated to be as narrow as possible, and as described above, in the positioning device 1, the area estimating unit 103 performs this narrowing down.

[About positioning using PDR (Pedestrian Dead Reckoning)]
The positioning method of the position estimating unit 105 is not limited to using magnetism. For example, the position estimation unit 105 may perform positioning using PDR. When performing positioning using PDR, the object of positioning (for example, person B in FIG. 3) may be equipped with a measuring device for measuring various data required for positioning using PDR. For example, an acceleration sensor or a gyro sensor may be worn on the body. Then, the position estimating unit 105 estimates the current position of the target by sequentially updating the position of the target using the measured time-series acceleration data and the like. Even when performing positioning using PDR, in order to improve the estimation accuracy by the position estimating unit 105, it is important to narrow down the area to be estimated to be as narrow as possible.

[Processing flow]
The flow of the process (positioning method) executed by the positioning device 1 will be explained based on FIG. 5. FIG. 5 is a flowchart illustrating an example of a process executed by the positioning device 1. In the following, an example will be described in which the measuring device 3 shown in FIG. 3 measures the atmospheric pressure, the received signal strength of LPWA, and magnetism, and the position of the person B who carries the measuring device 3 is determined from these measured values.

In S11, the data acquisition unit 101 acquires positioning data 111. The positioning data 111 includes data indicating the atmospheric pressure measured by the measuring device 3, the received signal strength of the LPWA, and magnetism. The positioning data 111 also includes data indicating the atmospheric pressure observed at a fixed point at a predetermined reference position (for example, the entrance of the first floor of building A). Note that the measurement of the atmospheric pressure at a predetermined reference position may be performed, for example, at predetermined time intervals, and when the atmospheric pressure is newly measured, the measured value may be added to the positioning data 111 along with the measurement time. Furthermore, the received signal strength of Wi-Fi or BLE may be acquired instead of LPWA.

In S12, the preprocessing unit 102 performs preprocessing on the positioning data 111 acquired in S11. As described above, the preprocessing unit 102 performs a process of subtracting the absolute error of the measuring device 3 as an offset from the atmospheric pressure value shown in the positioning data 111, a process of calculating the atmospheric pressure difference, and a process of calculating missing values of the received signal strength of the LPWA. Processing for interpolation, processing for scaling the atmospheric pressure difference and the interpolated LPWA reception signal strength, etc. are performed. Regarding the calculation of the atmospheric pressure difference, if the atmospheric pressure at the reference position is measured at predetermined intervals, the preprocessing unit 102 measures the atmospheric pressure at the reference position at the closest time to the time when the measuring device 3 measures the atmospheric pressure. All you have to do is calculate the difference between the atmospheric pressure and the calculated atmospheric pressure.

In S13 (area estimation step), the area estimation unit 103 uses the positioning data 111 to determine the area in the building A where the measurement point where the positioning data 111 was measured exists (that is, the area where the person B is present). presume.

Specifically, the area estimation unit 103 inputs the LPWA reception signal strength and the atmospheric pressure difference preprocessed in S12 to the estimation model 112. As a result, the estimation model 112 outputs, for each of a plurality of areas where a measurement point may exist, the probability that the measurement point exists in the area.

For example, assume that the following six areas are set for building A in FIG.

(1) Atrium area on the 1st floor (inside room A1 on the 1st floor)
(2) Area on the first floor that is not an atrium (in room A2 or A3)
(3) 2nd floor atrium area (inside room A1 on 2nd floor)
(4) Area that is not an atrium on the second floor (inside room A4)
(5) 3rd floor atrium area (inside room A1 on 3rd floor)
(6) Area on the third floor that is not an atrium (in room A5 or A6)
In this case, an estimate is generated by machine learning using teacher data 212 that labels the area where these are measured as correct data for the atmospheric pressure difference calculated from the atmospheric pressure measured in each area and the received signal strength of the LPWA. Model 112 may be used. If the received signal strength of the LPWA preprocessed in S12 and the atmospheric pressure difference are input to such an estimation model 112, the probability that a measurement point exists in each of the six areas described above is output.

In S14, the area integration unit 104 integrates a predetermined number of areas with higher probabilities estimated in S13 to form an integrated area. In S14, it is preferable to select areas to be integrated so that the sum of the estimated probabilities becomes 1 or a value close to 1. Note that the process in S14 may be omitted. In this case, in S13, the area with the highest estimated probability may be used as the estimation result.

In S15 (position estimation step), the position estimation unit 105 performs positioning targeting the integrated area. Specifically, the position estimating unit 105 selects the magnetism of the measurement point shown in the positioning data 111 acquired in S11 and the magnetism measured within the integrated area among the magnetisms shown in the verification magnetic data 113. The location of the measurement point is estimated by comparing the Note that if the process in S14 is omitted, in S15 the position of the measurement point in the area estimated in S13 is estimated. Further, the position estimating unit 105 may output the position estimation result to the output unit 14 or the like. With this, the process in FIG. 5 ends. Note that the process in FIG. 5 may be repeated at predetermined time intervals. Thereby, time-series positioning results of person B can be obtained.

As described above, the positioning method according to the present embodiment includes the area estimation step (S13) of estimating the area where the measurement point where the positioning data 111 was measured in the building exists, using the positioning data 111. , a position estimation step (S15) of estimating the position of the measurement point in the estimated area. As a result, positioning can be performed after narrowing down the positioning target area, so positioning can be performed with stable accuracy.

[Processing flow]
The flow of the process (learning method) executed by the learning device 2 will be explained based on FIG. 6. FIG. 6 is a flowchart illustrating an example of a process executed by the learning device 2. In the following, an example will be described in which an estimation model 112 is generated that estimates which of the above-mentioned six areas the measurement position corresponds to in building A shown in FIG. 3 based on the atmospheric pressure difference and the received signal strength of the LPWA.

In S21, the data acquisition unit 201 acquires learning data 211. The learning data 211 includes data indicating the atmospheric pressure measured in each of the six areas described above and the received signal strength of the LPWA. The learning data 211 also includes data indicating the atmospheric pressure observed at a fixed point at a predetermined reference position (for example, the entrance of the first floor of building A).

In S22, the teacher data generation unit 202 performs preprocessing on the learning data 211 acquired in S21. As described above, the teacher data generation unit 202 performs processes such as subtracting the absolute error of the sensor that measured the atmospheric pressure as an offset from the atmospheric pressure value shown in the learning data 211, calculating the atmospheric pressure difference, and calculating the received signal strength of the LPWA. Processing to interpolate missing values, processing to scale the atmospheric pressure difference and the interpolated LPWA reception signal strength, etc. are performed.

In S23, the teacher data generation unit 202 associates the received signal strength and the area where the atmospheric pressure was measured with the preprocessed air pressure difference and the received signal strength of the LPWA as a correct label, and generates the teacher data 212. generate. The above-mentioned correct labels for the six areas can be said to indicate the hierarchy of each point in building A where learning data was measured and whether or not the point is an atrium area. Further, the teacher data generation unit 202 may generate the teacher data 212 by adding a noise component to the received signal strength of the LPWA.

In S24 (teacher data acquisition step), the teacher data acquisition unit 203 acquires the teacher data 212 generated in S23. The training data 212 in this example includes the difference between the atmospheric pressure measured at each point in building A and the atmospheric pressure measured at a predetermined reference point, and the received signal strength of the LPWA measured at each point, and the hierarchy of each point. It can be said that whether or not the point is in an atrium area is associated with the correct label.

In S25 (learning step), the learning unit 204 generates the estimated model 112 by learning using the teacher data 212 acquired in S24. As described above, the teacher data 212 can be said to be the one in which the atmospheric pressure difference and the received signal strength of the LPWA are associated with the hierarchy of each point and whether or not the point is in an atrium area as a correct label. Therefore, the estimation model 112 generated by learning using the teacher data 212 is based on the atmospheric pressure difference between the measurement point and the reference point and the received signal strength of the LPWA measured at the measurement point, and the hierarchy of the measurement point and the It can be said that this is a model for estimating whether a point is in an atrium area or not.

As described above, the learning method according to the present embodiment is based on the difference between the atmospheric pressure measured at each point in the building and the atmospheric pressure measured at a predetermined reference point and the received signal strength of the LPWA. and a teacher data acquisition step (S24) in which teacher data 212 is associated with whether or not the point is an atrium area as a correct answer label, and learning using the teacher data 212 is performed to identify the measurement point and the reference point. A learning step of generating an estimation model 112 for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area from the pressure difference and the received signal strength of the LPWA measured at the measurement point ( S25). Thereby, the estimation model 112 that can stably perform area estimation can be generated. Note that it is not essential to use the received signal strength of the LPWA, and the estimation model 112 that estimates the area only from the pressure difference may be generated.

[Modified example]
The above-mentioned area setting is only an example, and it is arbitrary what kind of area is set within the building that is the positioning target. For example, in a building including multiple floors, one floor may be set as one area. Furthermore, the positioning device 1 can also be used for positioning in a building without an atrium.

Further, the execution entity of each process described in the above-described embodiment is arbitrary and is not limited to the above-mentioned example. In other words, the function of the positioning device 1 can be replaced by a plurality of information processing devices (which can also be called processors) that can communicate with each other. For example, by distributing and providing each block shown in FIG. 1 in a plurality of information processing devices, a system having the same functions as the positioning device 1 can be constructed. Therefore, the main body that executes the positioning method shown in FIG. 5 can also be a plurality of information processing apparatuses. The same applies to the learning device 2, and the functions of the learning device 2 can be replaced by multiple information processing devices that can communicate with each other, and the learning method shown in FIG. 6 is also executed by the multiple information processing devices. It can be done.

[Example of implementation using software]
The function of the positioning device 1 and the learning device 2 (hereinafter referred to as "device") is a program for making a computer function as the device, and each control block of the device (particularly included in the control units 10 and 20) Each part) can be realized by a program (positioning program/learning program) for making the computer function.

In this case, the device includes a computer having at least one control device (for example, a processor) and at least one storage device (for example, a memory) as hardware for executing the program. By executing the above program using this control device and storage device, each function described in each of the above embodiments is realized.

The above program may be recorded on one or more computer-readable recording media instead of temporary. This recording medium may or may not be included in the above device. In the latter case, the program may be supplied to the device via any transmission medium, wired or wireless.

Further, part or all of the functions of each of the control blocks described above can also be realized by a logic circuit. For example, an integrated circuit in which a logic circuit functioning as each of the control blocks described above is formed is also included in the scope of the present invention. In addition to this, it is also possible to realize the functions of each of the control blocks described above using, for example, a quantum computer.

〔summary〕
The positioning device according to aspect 1 of the present invention includes an area estimating unit that estimates an area in which a measurement point where positioning data was measured exists in a building using the positioning data; A position estimation unit that estimates the position of the measurement point.

In the positioning device according to aspect 2 of the present invention, in aspect 1, the area estimating unit calculates the probability that the measurement point exists in the area for each of a plurality of areas where the measurement point may exist. and the position estimating unit estimates the position of the measurement point in an integrated area that integrates a predetermined number of areas with higher estimated probabilities.

In the positioning device according to aspect 3 of the present invention, in aspect 1 or 2, the building is a multi-story building including an atrium area and a non-atrium area, and the positioning data includes the Data indicating the atmospheric pressure measured at the measuring point is included, and the area estimating section calculates the atmospheric pressure at the measuring point based on the difference between the atmospheric pressure measured at the measuring point and the atmospheric pressure measured at a predetermined reference point. , and whether or not the measurement point is in an atrium area.

In the positioning device according to aspect 4 of the present invention, in any one of aspects 1 to 3, the positioning data includes received signal strength of LPWA (Low Power Wide Area-network) measured at the measurement point. The area estimator estimates the area where the measurement point is located based on the received signal strength.

In the learning device according to aspect 5 of the present invention, the difference between the atmospheric pressure measured at each point in the building and the atmospheric pressure measured at a predetermined reference point is determined by the floor level of each point and the area of the atrium. A teacher data acquisition unit that acquires teacher data in which whether or not is associated as a correct answer label, and learning using the teacher data, determine the hierarchy of the measurement point and the measurement point from the pressure difference between the measurement point and the reference point. and a learning unit that generates an estimation model for estimating whether or not the area is an atrium area.

In the learning device according to aspect 6 of the present invention, in the aspect 5, the teacher data includes data indicating the received signal strength of the LPWA measured at each point in the building, and the received signal of the LPWA is A teacher data generating section is provided that generates teacher data using data obtained by adding a noise component to the data indicating the intensity.

A positioning method according to aspect 7 of the present invention is a positioning method executed by one or more information processing devices, in which an area where a measurement point where positioning data was measured is located in a building is The method includes an area estimating step of estimating using data, and a position estimating step of estimating the position of the measurement point in the estimated area.

A learning method according to aspect 8 of the present invention is a learning method executed by one or more information processing devices, in which the atmospheric pressure measured at each point in a building is compared with the atmospheric pressure measured at a predetermined reference point. A step of acquiring teacher data in which the hierarchy of each point and whether or not the point is in an atrium area is associated with the difference as a correct label; and learning using the teacher data to determine the measurement point. and a learning step of generating an estimation model for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area based on the pressure difference between the measurement point and the reference point.

A positioning program according to aspect 9 of the present invention is a positioning program for causing a computer to function as the positioning device according to aspect 1, and causes the computer to function as the area estimation section and the position estimation section.

The present invention is not limited to the embodiments described above, and various modifications can be made within the scope of the claims, and embodiments can be obtained by appropriately combining technical means disclosed in different embodiments. are also included within the technical scope of the present invention.

1 Positioning device 103 Area estimation unit 105 Position estimation unit 111 Data for positioning 112 Estimation model 2 Learning device 202 Teacher data generation unit 203 Teacher data acquisition unit 204 Learning unit 212 Teacher data

Claims (9)

an area estimating unit that estimates an area where a measurement point from which positioning data was measured in the building exists, using the positioning data;
A positioning device comprising: a position estimation unit that estimates a position of the measurement point in the estimated area. The area estimation unit estimates, for each of a plurality of areas where the measurement point may exist, a probability that the measurement point exists in the area;
The positioning device according to claim 1, wherein the position estimation unit estimates the position of the measurement point in an integrated area that integrates a predetermined number of areas with higher estimated probabilities. The building is a multi-story building including an atrium area and a non-atrium area,
The positioning data includes data indicating the atmospheric pressure measured at the measurement point,
The area estimation unit determines the hierarchy of the measurement point and whether or not the measurement point is an atrium area based on the difference between the atmospheric pressure measured at the measurement point and the atmospheric pressure measured at a predetermined reference point. The positioning device according to claim 1 or 2, which estimates . The positioning data includes data indicating a received signal strength of an LPWA (Low Power Wide Area-network) measured at the measurement point,
The positioning device according to claim 1 or 2, wherein the area estimator estimates an area where the measurement point exists based on the received signal strength. The difference between the atmospheric pressure measured at each point in the building and the atmospheric pressure measured at a predetermined reference point is associated with the hierarchy of each point and whether or not the point is an atrium area as a correct answer label. a teacher data acquisition unit that acquires teacher data;
a learning unit that generates an estimation model for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area from the pressure difference between the measurement point and the reference point by learning using the teacher data; A learning device comprising: The training data includes data indicating the received signal strength of the LPWA measured at each point in the building,
The learning device according to claim 5, further comprising a teacher data generation unit that generates teacher data using data obtained by adding a noise component to the data indicating the received signal strength of LPWA. A positioning method executed by one or more information processing devices, the method comprising:
an area estimation step of estimating an area where a measurement point where positioning data was measured in the building exists, using the positioning data;
A positioning method comprising: estimating a position of the measurement point in the estimated area. A learning method executed by one or more information processing devices,
The difference between the atmospheric pressure measured at each point in the building and the atmospheric pressure measured at a predetermined reference point is associated with the hierarchy of each point and whether or not the point is an atrium area as a correct answer label. a teacher data acquisition step of acquiring teacher data;
a learning step of generating, by learning using the teacher data, an estimation model for estimating the hierarchy of the measurement point and whether or not the measurement point is in an atrium area from the pressure difference between the measurement point and the reference point; , learning methods including; A positioning program for causing a computer to function as the positioning device according to claim 1, the positioning program for causing the computer to function as the area estimation section and the position estimation section.

Images