JP7410510B2 - Location estimation device, location estimation method, and location estimation program - Google Patents

Description

The present invention relates to a position estimation device, a position estimation method, and a position estimation program.

With the spread of user terminals such as mobile phones and smartphones, the number of services that utilize the location of the user terminals is increasing. The position of the user terminal is generally determined using signals from Global Positioning System (GPS) satellites. On the other hand, for example, it is assumed that signals from GPS satellites cannot be properly received inside facilities such as shopping malls, airports, and stations.

Therefore, information indicating the strength of a signal received using a wireless Local Area Network (LAN) (also referred to as Wi-Fi (registered trademark)) (for example, IEEE802.11n, IEEE802.11ac, etc.) (received signal strength information ) have been studied to estimate the position of a user terminal (for example, Patent Document 1). In Patent Document 1, the position of a user terminal within a facility is estimated using the RSSI of a signal from the user terminal received at a wireless LAN access point (AP).

Japanese Patent Application Publication No. 2018-205249

In recent years, wireless LAN access points have been installed in facilities such as shopping malls, airports, and stations. Further, user terminals are generally equipped with a communication function using a wireless LAN. Therefore, position estimation of a user terminal using received signal strength information in a wireless LAN can utilize existing equipment, and installation costs can be suppressed. However, the coverage area of the access point is limited to a radius of several meters, and obstacles may occur between the access point and the user terminal. Therefore, there is a problem in that the position of the user terminal estimated based on the received signal strength information in the wireless LAN often contains errors.

Therefore, one of the objects of the present invention is to provide a position estimating device, a position estimating method, and a position estimating program that can improve the accuracy of estimating the position of a user terminal based on received signal strength information in a wireless LAN.

A position estimating device according to an aspect of the present invention corresponds to received signal strength information indicating the strength of a signal from a predetermined device at one or more access points and position information indicating the position of the predetermined device within a predetermined area. a storage unit that stores a position estimation model generated by machine learning using the attached learning data; and an acquisition unit that acquires received signal strength information indicating the strength of a signal from the user terminal at the one or more access points. , an estimation unit that inputs the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area.

According to this aspect, the received signal strength information indicating the strength of the signal from the user terminal at one or more access points is input into the position estimation model, and the position of the user terminal within the predetermined area is estimated. The accuracy of estimating the location of a user terminal based on received signal strength information can be improved.

In the above aspect, the received signal strength information and the position in the learning data are determined based on whether the date and time associated with the received signal strength information and the date and time associated with the location information are within a predetermined time range. An association with information may also be determined.

According to this aspect, since the learning data used for generating the position estimation model can be appropriately generated, the accuracy of estimating the position of the user terminal that is estimated by inputting the received signal strength information to the position estimation model is improved. can.

In the above aspect, the received signal strength information and the location in the learning data are determined based on whether the terminal identification information associated with the received signal strength information and the terminal identification information associated with the location information are the same. An association with information may also be determined.

According to this aspect, since the learning data used for generating the position estimation model can be appropriately generated, the accuracy of estimating the position of the user terminal that is estimated by inputting the received signal strength information to the position estimation model is improved. can.

In the above aspect, the signal from the predetermined device and/or the user terminal is a probe request transmitted from the predetermined device and/or the user terminal, or a connection is established between the predetermined device and each access point. The data signal may be a data signal transmitted from the predetermined device after being transmitted. According to this aspect, since the probe request is transmitted from any terminal equipped with a wireless LAN communication function, the acquisition of received signal strength information can be made more efficient.

In the above aspect, the position information may be a point of two-dimensional coordinates or three-dimensional coordinates estimated using a sensor mounted on the predetermined device. According to this aspect, it is possible to efficiently generate learning data used to generate a position estimation model.

The above aspect may further include an output unit that outputs information indicating a flow line of the user terminal based on the estimated position of the user terminal. According to this aspect, the flow line of the user terminal can be grasped with high precision while using existing access points provided in a predetermined area.

A position estimation method according to another aspect of the present invention includes received signal strength information indicating the strength of a signal from a predetermined device at one or more access points, and position information indicating the position of the predetermined device within a predetermined area. A location estimation method using a location estimation model generated by machine learning using associated learning data, wherein received signal strength information indicating the strength of a signal from a user terminal at the one or more access points is obtained. and inputting the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area.

A position estimation program according to another aspect of the present invention includes received signal strength information indicating the strength of a signal from a predetermined device at one or more access points, and position information indicating the position of the predetermined device within a predetermined area. A calculation unit equipped with a position estimation device that stores a position estimation model generated by machine learning using the associated learning data receives a received signal indicating the strength of a signal from the user terminal at the one or more access points. It functions as an acquisition unit that acquires strength information, and an estimation unit that inputs the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area.

According to the present invention, it is possible to improve the accuracy of estimating the position of a user terminal based on received signal strength information in a wireless LAN.

1 is a diagram illustrating an example of a schematic configuration of a position estimation system according to an embodiment. FIG. 1 is a diagram illustrating an example of the functional configuration of a position estimation device and a learning device according to the present embodiment. It is a figure showing an example of RSSI information concerning this embodiment. It is a figure showing an example of position related information concerning this embodiment. It is a figure showing an example of data for learning concerning this embodiment. It is a diagram showing an example of the hardware configuration of each device in the position estimation system according to the present embodiment. It is a flowchart which shows an example of learning operation concerning this embodiment. It is a figure showing an example of generation of data for learning concerning this embodiment. It is a flowchart which shows an example of position estimation operation concerning this embodiment. FIG. 3 is a diagram showing an example of a flow line of a user terminal according to the present embodiment.

Embodiments of the present invention will be described with reference to the accompanying drawings. In addition, in each figure, those with the same reference numerals have the same or similar configurations.

(Configuration of position estimation system)
<Schematic configuration>
FIG. 1 is a diagram showing an example of a schematic configuration of a position estimation system 1 according to the present embodiment. The position estimation system 1 includes a position estimation device 10, a learning device 20, a user terminal 30, an access point (AP) 40 that communicates with the user terminal 30 using a wireless LAN, and a self-position estimation device 50. have Note that the wireless LAN is also called Wi-Fi (registered trademark), and may comply with, for example, IEEE802.11n, IEEE802.11ac, etc.

The position estimation device 10 is a device that estimates the position of the user terminal 30 using a position estimation model generated based on machine learning. Specifically, the position estimation device 10 uses information indicating the strength of signals from the user terminals 30 at N (N≧1) APs 40 _i (1≦i≦N, for example, N=3 in FIG. 1). (received signal strength information) is input into a position estimation model to estimate the position of the user terminal 30. The received signal strength information is also called received signal strength, received strength, received strength information, received signal strength indicator (RSSI), etc. Hereinafter, the received signal strength information will be referred to as RSSI, but the name is not limited to this.

The position estimating device 10 may be connected to the AP 40 _i via the network N. Further, the position estimation device 10 may be connected to the learning device 20 via the network N. Note that the network N may include, for example, one or more network devices such as a gateway. Further, the network N may include one or more networks (eg, the Internet, a wireless LAN, etc.).

The learning device 20 performs a learning process using machine learning of a position estimation model used to estimate the position of the user terminal 30. Specifically, the learning device 20 uses RSSI (received signal strength information) indicating the received signal strength from the self-position estimating device 50 measured by one or more APs 40 (for example, three APs 40) and the self-position estimating device. Based on the information (position information) indicating the position of the self-position estimating device 50 itself measured by the self-position estimating device 50, a learning process by machine learning of the position estimation model is performed.

The user terminal 30 is a terminal whose position within a given area is estimated using a position estimation model installed in the position estimation device 10, and is, for example, a smartphone, a personal computer (PC), or the like. The predetermined area may be, for example, an indoor commercial facility (e.g., shopping center, airport, train station, etc.) or a sports arena (e.g., indoor stadium, gymnasium, indoor stadium, swimming pool, etc.) that receives signals from GPS satellites. It may be a space where reception is difficult. Further, the predetermined area may be each room or each area within these facilities. Furthermore, the user terminal 30 has a communication function using the wireless LAN described above.

The AP 40 _i (1≦i≦N, for example, N=3 in FIG. 1) is an access point of the wireless LAN, and is also called a base station, wireless base station, access node, node, etc. The AP 40 _i (i≧1) forms a cell (communication area) C _i with a predetermined range (for example, a radius of about 10 meters). The AP 40 _i is located within the cell C and communicates with a terminal (for example, the user terminal 30, the self-position estimating device 50, etc.) that has a wireless LAN communication function.

The AP 40 _i may be connected to the position estimation device 10 and the learning device 20 via the network N. Note that although three APs 40 ₁ , 40 ₂ and 40 ₃ are shown in FIG. 1, the number is not limited to this, and any number of APs may be one or more. One or more APs 40 _i are provided within the predetermined area. Note that when one or more APs 40 _i (i≧1) and one or more cells C _i formed by the AP _i are not particularly distinguished from each other, they are collectively referred to as APs 40 and cells C.

The self-position estimation (localization) device 50 is a mobile object (for example, a cart, a wheeled trolley, a drone, etc.) that has a self-position estimation (localization) function for estimating the position of the self-position estimation device 50 itself. Further, the self-position estimating device 50 may be provided with a communication function using the wireless LAN, or may not be provided with a communication function using the wireless LAN. For example, the self-position estimating device 50 may be one in which the user terminal 30 is mounted on a cart equipped with the self-position estimating function. The self-position estimation function may be realized by, for example, at least one of the following sensors or functions.
・Simultaneous Localization And Mapping (SLAM)
・Laser sensor (e.g. Light Detection And Ranging (LiDAR))
・Radar sensor (e.g. RAdio Detection And Ranging (RADAR))
・Camera (Image) Sensor
・Extended Kalman Filter
・Unscented Kalman Filter
・Histogram Filter
・Particle Filter
・Gyro Sensor
・Accelerometer
・Indoor MEssaging System (IMES)
・Infrared sensor ・Global Positioning System (GPS)
・Global Navigation Satellite System (GNSS)

For example, if the self-position estimating device 50 is equipped with a SLAM using a laser sensor (also referred to as LiDAR SLAM, etc.), the output value of the laser sensor (for example, the position (x, y) on two-dimensional coordinates or the position (x, y) on three-dimensional coordinates) A map (for example, a two-dimensional or three-dimensional point cloud map, a grid map, or a voxel map) of the predetermined area may be constructed based on the point cloud data of the positions (x, y, z) above. . Specifically, the self-position estimating device 50 may sequentially estimate the amount of movement by matching point groups output from the laser sensors, and may estimate the self-position by integrating the amounts of movement.

In FIG. 1, the self-position estimating device 50 is not limited to a single device incorporating a self-position estimating function (or a self-position estimating function and a wireless LAN communication function), but may be configured by combining a plurality of devices. may be done. For example, the self-position estimating device 50 connects a device having a self-value estimation function (for example, a SLAM device) and a device having a wireless LAN communication function (for example, the user terminal 30) to a mobile object (for example, a cart, a wheel, etc.). It may also be configured by being mounted on a trolley, drone, etc.). Note that in the following, the self-position estimating device includes a moving object or the like carrying a SLAM device equipped with a predetermined sensor.

Further, although this example shows a case where the position estimation device 10 and the learning device 20 are separate bodies, the position estimation device 10 and the learning device 20 may be configured as an integrated device. For example, the position estimation device 10 and the learning device 20 may be realized as different operations of PLC (Programmable Logic Controller).

<Functional configuration>
≪Learning device≫
FIG. 2 is a diagram illustrating an example of the functional configuration of a position estimation device and a learning device according to this embodiment. As shown in FIG. 2, the learning device 20 includes an RSSI information acquisition section 21, a location-related information acquisition section 22, a generation section 23, a storage section 24, and a learning section 25.

The RSSI information acquisition unit 21 acquires information regarding RSSI (RSSI information) indicating the strength of a signal from the self-position estimating device 50 (predetermined device) in the AP 40 . The signal from the self-position estimating device 50 is, for example, transmitted from the self-position estimating device 50 (the communication function of the wireless LAN that constitutes the self-position estimating device 50 or the user terminal that constitutes the self-position estimating device 50). It may be a probe request that is transmitted, or a data signal that is transmitted from the self-position estimating device 50 after the connection between the self-position estimating device 50 and each AP 40 is established.

Here, the probe request (also referred to as a probe request frame or the like) is a signal used by a wireless LAN terminal to search for nearby APs 40 (active scan). The probe request includes (1) identification information of the terminal sending the probe request (also referred to as terminal identification information, source address (SA), etc.) (for example, MAC (Media Access Control) address); (2) It may include at least one sequence number. The sequence number may be counted up each time a probe request is transmitted at each terminal, and the same sequence number may indicate the same probe request. Further, the data signal may include at least one of the terminal identification information and the sequence number.

Specifically, the RSSI information acquisition unit 21 obtains RSSI, identification information (terminal identification information) of the self-position estimating device 50 that transmits the RSSI measurement signal, and identification information (AP ID) of the AP 40 that measured the RSSI. , RSSI information that associates the date and time related to the RSSI (for example, RSSI measurement date and time or reception date and time, also referred to as a timestamp) may be obtained from each AP 40. Furthermore, in addition to the RSSI, terminal identification information, and date and time regarding the RSSI, the RSSI information may be associated with a sequence number of a signal for measuring the RSSI.

FIG. 3 is a diagram showing an example of RSSI information according to this embodiment. Note that in FIG. 3, the signal for measuring RSSI is the probe request, but the signal is not limited to this, and may be the data signal. As shown in FIG. 3, the RSSI information includes the RSSI of a signal from the self-position estimating device 50 (for example, a probe request), the terminal identification information of the self-position estimating device 50 (for example, the MAC address included in the probe request), The AP-ID of the AP 40 where the RSSI is measured, the sequence number of the probe request, and the date and time regarding the RSSI (for example, the date and time of RSSI measurement) may be associated.

As mentioned above, the probe request sequence number is counted up each time a probe request is sent by the terminal. Therefore, multiple RSSIs associated with the same terminal identification information and the same sequence number are determined to be multiple RSSIs measured by different APs 40 based on the same probe request.

For example, in FIG. 3, each RSSI associated with sequence number "001" may be determined to have been measured by each AP 40 based on a probe request transmitted from the self-position estimating device 50 at a point A within a predetermined area. . Further, each RSSI associated with the sequence number "002" is transmitted from the self-position estimating device 50 at point B within the predetermined area after a predetermined period (for example, 60 seconds) from the probe request with the sequence number "001". It may be determined that each AP 40 has measured based on the probe request. A similar determination may be made for each RSSI associated with sequence number "003". The distances between the self-position estimating device 50 and each AP 40 at the probe request transmission positions (for example, points A, B, and C) within the predetermined area are not the same, and there may be obstacles. Therefore, as shown in FIG. 3, multiple RSSIs measured at different APs 40 based on the same probe request may have different values.

In this way, whether multiple RSSIs associated with different AP IDs are based on the same probe request may be determined based on the sequence numbers associated with each of the multiple RSSIs. Alternatively, whether the plurality of RSSIs are based on the same probe request may be determined based on the date and time respectively associated with the plurality of RSSIs. For example, if multiple dates and times associated with the same terminal identification information to the multiple RSSIs are within a predetermined time range (for example, 1 second), if the multiple RSSIs are measured based on the same probe request. You may be judged. The RSSI information acquisition unit 21 may group one or more RSSIs that are determined to be based on the same probe request.

Similarly, when using the above data signal to measure RSSI in each AP 40, one or more RSSIs based on the same data signal may be grouped. The grouping may be performed based on the date and time associated with each RSSI using the same terminal identification information, or may be performed based on the sequence number within the data signal. For example, if multiple dates and times associated with the same terminal identification information to the multiple RSSIs are within a predetermined time range (for example, 1 second), if the multiple RSSIs are measured based on the same data signal. You may be judged.

Note that if one or more RSSIs based on the same probe request or data signal are grouped based on date and time, the RSSI information does not need to include a sequence number. That is, in the RSSI information, the RSSI, the terminal identification information of the self-position estimating device 50, the AP-ID of the AP 40 where the RSSI is measured, and the date and time regarding the RSSI may be associated.

Note that in FIG. 3, the RSSI information is RSSI measured by five APs 40 ₁ to 40 ₅ based on the same probe request, but the RSSI is measured based on the same probe request or the above data signal. The number of APs 40 (ie, the number of RSSIs measured based on the same probe request or data signal) is not limited to three. RSSI measured based on the probe request or the data signal transmitted from the self-position estimating device 50 or the number of APs 40 existing within the reach range (for example, about 10 meters) of the data signal. The number of may vary.

Further, although not shown, the RSSI information may include RSSI measured at each AP 40 within the predetermined area, regardless of the probe request or the detection of the data signal. An RSSI indicating "null" or "0" may be associated with the AP ID of the AP 40 that has not detected the probe request or the data signal.

Further, the RSSI information may include RSSIs that satisfy a predetermined condition (or the RSSI information acquisition unit 21 may extract RSSIs that satisfy a predetermined condition). The predetermined condition is, for example, that the RSSI value is greater than or equal to a predetermined threshold, or that the RSSI value is equal to or greater than a predetermined threshold or a predetermined number of values (for example, 3) RSSI.

The position-related information acquisition unit 22 acquires information (position-related information) regarding the position of the self-position estimating device 50 (predetermined device) within a predetermined area. The location-related information includes information indicating the position of the self-position estimating device 50 within a predetermined area (position information), identification information of the self-position estimating device 50 (terminal identification information), and the date and time related to the position information (for example, the location information). (estimated date and time of the information). The position information indicates the own position estimated by the self-position estimating device 50, and includes, for example, a predetermined sensor (for example, a laser sensor, various radio wave receivers (for example, GPS, GNSS, etc.) mounted on the self-position estimating device 50). ), acceleration sensor, gyro sensor, image sensor) (for example, a position on two-dimensional coordinates (x, y) or a position on three-dimensional coordinates (x, y, z)). .

FIG. 4 is a diagram illustrating an example of location-related information according to this embodiment. Note that in FIG. 4, it is assumed that the position information included in the position-related information is information (x, y) indicating a two-dimensional coordinate position output from a predetermined sensor of the self-position estimating device 50; The information is not limited to this, and may be any information estimated by the self-position estimation function.

As shown in FIG. 4, the position-related information includes position information indicating the position of the self-position estimating device 50 within a predetermined area (in FIG. 4, the position on two-dimensional coordinates), and the date and time related to the position information (for example, the position (estimated date and time of the information) may also be associated. Although not shown, the location-related information may be associated with identification information of a predetermined area (eg, room, area, sales floor, floor, etc.) that includes the location indicated by the location information. By associating the identification information with position information, a detailed position within a predetermined area (for example, a certain room) can be grasped. Further, terminal identification information (for example, MAC address) of the self-position estimating device 50 may be associated with the location information and date and time.

The generation unit 23 generates learning data based on the RSSI information acquired by the RSSI information acquisition unit 21 and the location-related information acquired by the location-related information acquisition unit 22. Specifically, the generation unit 23 performs learning that associates the RSSI in each AP 40 in the predetermined area with the location information based on the measurement date and time in the RSSI information and the estimated date and time in the location-related information. You may also generate data for

For example, the generation unit 23 generates a signal that is measured based on the same signal (for example, the same probe request) in the learning data based on the measurement date and time associated with the RSSI and the estimated date and time associated with the location information. The correspondence between the above RSSI and the location information of the self-position estimating device 50 may be determined (the one or more RSSIs and the location information may be compared or matched).

FIG. 5 is a diagram showing an example of learning data according to this embodiment. In the learning data of FIG. 5, one or more of the RSSIs of FIG. The correspondence between the RSSI of and the location information of FIG. 4 is determined. In FIG. 5, it is assumed that 11 APs 40 ₁ to 40 ₁₁ are provided within the predetermined region whose position is to be estimated, but the number of APs 40 within the predetermined region is not limited to this.

For example, the time range determined based on the estimated date and time of the location information in FIG. 21.000 seconds), there are five RSSIs measured by the APs 40 with AP-IDs "1" to "5" in FIG. In this case, as shown in FIG. 5, in the learning data, the estimated date and time of the location information "July 7, 2019 8:19 minutes 20.254 seconds" and the RSSI of AP40 ₁ to 40 ₅ are associated. It's okay to be hit.

Here, during the time range (for example, 8:19:20,000 to 21,000 seconds on July 7, 2019), in FIG. 3, AP40 ₆ with AP-ID "6" to "11" 40 ₁₁ is not included (ie, no probe request or data signals are detected at APs 40 ₆ to 40 ₁₁ in that time range). Therefore, as shown in FIG. 5, in the learning data, the RSSI of AP40 ₆ to 40 ₁₁ that is associated with the location information of the estimated date and time of "July 7, 2019 8:19 minutes 20.254 seconds" may be set to 0 or null. In this way, the learning data may be information that associates the RSSI and position information of each AP 40 that are respectively associated with the measurement date and time and the estimated date and time that exist within the same time range.

The storage unit 24 stores the learning data generated by the generation unit 23 and the position estimation model.

The learning unit 25 executes learning processing of the position estimation model by machine learning using the learning data stored in the storage unit 24. Specifically, the learning unit 25 receives as input the RSSI measured by one or more APs 40 in the learning data, and performs a learning process on the position estimation model so that the position information associated with the RSSI is output. It's okay.

The position estimation model subjected to machine learning by the learning unit 25 may be configured by, for example, a neural network. When the position estimation model is configured by a neural network, the learning unit 25 may perform learning processing of the neural network using, for example, an error backpropagation method. Also. The position estimation model may be configured based on deep learning, which is a neural network having two or more intermediate layers.

≪Position estimation device≫
As shown in FIG. 2, the position estimation device 10 includes an RSSI information acquisition section 11, an estimation section 12, a storage section 13, and a control section 14.

The RSSI information acquisition unit 11 acquires RSSI information related to the RSSI of a signal (for example, a probe request) from the user terminal 30, which is measured by each AP 40. Specifically, the RSSI information acquisition unit 11 acquires the RSSI, the identification information (terminal identification information) of the user terminal 30 that transmits the RSSI measurement signal, the identification information (AP ID) of the AP 40 that measured the RSSI, the RSSI information that associates dates and times regarding RSSI may be acquired from each AP 40.

As explained in the RSSI information acquisition unit 21, the RSSI information acquisition unit 11 determines whether multiple RSSIs associated with different AP IDs in the RSSI information are the same signal from the user terminal 30 (for example, the same probe request or the same data signal) may be determined based on the sequence number or measurement date and time respectively associated with the plurality of RSSIs. The RSSI information acquisition unit 11 may group one or more RSSIs that are determined to be based on the same signal. The details of the RSSI information acquisition section 11 are as described for the RSSI information acquisition section 21.

The estimation unit 12 inputs the RSSI indicating the strength of the signal from the user terminal 30 at one or more APs 40 into the position estimation model, and estimates the position of the user terminal 30 within a predetermined area. Specifically, the estimation unit 12 inputs one or more RSSIs grouped by the RSSI information acquisition unit 11 into a position estimation model, and estimates the position of the user terminal 30 . The estimation unit 12 may cause the storage unit 13 to store information indicating the estimated position (estimated position information).

The position estimation model implemented in the estimation unit 12 may be a position estimation model generated by machine learning in the learning device 20, or may be a position estimation model generated by "distillation" in machine learning or a position estimation model generated by distillation. It may be a new position estimation model. In "distillation", a new prediction model may be generated using input/output data to the position estimation model machine-learned by the learning device 20. Note that the position estimation model may be expressed as a position estimation program, a position estimation algorithm, an estimator, a prediction program, a prediction algorithm, a prediction model, a program, an algorithm, a model, etc.

The control unit 14 controls the output of information indicating the flow line of the user terminal (flow line information) based on the position of the user terminal 30 estimated by the estimation unit 12. Specifically, the control unit 14 causes the output unit 10e (for example, a display) to display flow line information of the user terminal 30 generated based on the estimated position information stored in the storage unit 13, or causes the communication unit It may be transmitted to another device via 10c.

<Hardware configuration>
Next, each device in the position estimation system 1 (for example, the position estimation device 10, the learning device 20, the user terminal 30, the AP 40, the self-position estimation device 50, and at least one of the devices constituting the self-position estimation device 50) Explain the hardware configuration. As shown in FIG. 6, each device in the position estimation system 1 includes a CPU (Central Processing Unit) 10a corresponding to an arithmetic device, a storage device 10b, a communication section 10c, an input section 10d, and an output section 10e. has. These components are connected to each other via a bus so that they can transmit and receive data. Note that in this example, the position estimation device 10 and the learning device 20 are configured as separate computers, but may be configured as a single computer.

The CPU 10a is a control unit that performs control related to the execution of programs stored in the storage device 10b, and performs calculations and processing of data. The CPU 10a may be a calculation device (calculation unit) that executes a program (position estimation program) that estimates the position of the user terminal 30 based on the RSSI of the signal from the user terminal 30 in one or more APs 40. The CPU 10a receives various input data from the input section 10d and/or the communication section 10c, and outputs (for example, displays) the calculation result of the input data to the output section 10e, stores it in the storage device 10b, or performs communication. The data may be transmitted via the unit 10c.

The storage device 10b is at least one of a memory, an HDD (Hard Disk Drive), and an SSD (Solid State Drive). The storage device 10b of the learning device 20 may constitute the storage section 24. The storage device 10b of the position estimation device 10 may store a position estimation program executed by the CPU 10a. Further, the storage device 10b of the position estimation device 10 may constitute the storage section 13.

The communication unit 10c is an interface that connects each device in the position estimation system 1 to external equipment. Note that when the position estimation device 10 and the learning device 20 are configured as an integrated device, the communication unit 10c performs inter-process communication between the process that operates as the position estimation device 10 and the process that operates as the learning device 20. may be included.

The input unit 10d receives data input from the user, and may include, for example, at least one of a keyboard, a mouse, a touch panel, and a microphone.

The output unit 10e outputs the calculation results by the CPU 10a, and may include, for example, at least one of a display such as an LCD (Liquid Crystal Display) and a speaker.

The position estimation program may be provided by being stored in a computer-readable storage medium such as the storage device 10b, or may be provided via the network N connected by the communication unit 10c. The storage medium storing the determination program may be a non-transitory computer readable medium. The non-temporary storage medium is not particularly limited, and may be, for example, a USB memory, a CD-ROM, or a DVD.

In the position estimation device 10, the CPU 10a executes the position estimation program, thereby realizing the operations of the RSSI information acquisition section 11, the estimation section 12, and the control section 14 described using FIG. Note that these physical configurations are merely examples, and do not necessarily have to be independent configurations. For example, the position estimation device 10 may include an LSI (Large-Scale Integration) in which the CPU 10a and the storage device 10b are integrated.

Further, the RSSI information acquisition units 11 and 21 and the location-related information acquisition unit 22 may acquire the RSSI information and location-related information received by the communication unit 10c, respectively, or by using a computer such as the storage device 10b. RSSI information and location-related information stored on a readable storage medium may be obtained.

(Operation of position estimation system)
<Learning movement>
FIG. 7 is a flowchart showing an example of the learning operation according to this embodiment. As shown in FIG. 7, in step S101, the learning device 20 uses the signal transmitted from the self-position estimating device 50 to acquire the RSSI measured by each AP 40. Specifically, the learning device 20 may acquire RSSI information (for example, FIG. 3) including the RSSI from each AP 40. Further, the learning device 20 may group RSSIs measured by one or more APs 40 based on the same signal from the self-position estimating device 50.

In step 102, the learning device 20 acquires position information indicating the position of the self-position estimating device 50. Specifically, the learning device 20 may acquire location-related information (for example, FIG. 4) including the location information from the self-position estimation device 50.

In step S103, the learning device 20 generates learning data in which one or more RSSIs acquired in step S101 are associated with the position information of the self-position estimation device 50 acquired in step S102.

In step S104 in FIG. 7, the learning device 20 executes a position estimation model learning process by machine learning using the learning data generated as described above.

FIG. 8 is a diagram illustrating an example of a movement route of the self-position estimating device in a learning operation according to the present embodiment. In FIG. 8, it is assumed that the self-position estimating device 50 moves around a predetermined area (for example, a target area in a shopping mall) and transmits a probe request at a predetermined period (for example, 60 seconds). Furthermore, in FIG. 8, it is assumed that the self-position estimating device 50 performs self-position estimation and creation of an environmental map using SLAM. Note that self-position estimation may be performed at a period equal to or shorter than the period of the probe request, in conjunction with the transmission timing of the probe request. Note that in FIG. 8, it goes without saying that the above data signal may be used instead of the probe request.

Further, in FIG. 8, it is assumed that RSSI is measured using a probe request from the self-position estimating device 50, but as described above, the signal for measuring RSSI is not limited to the probe request. Further, although in FIG. 8, the same probe request is identified by the sequence number of the probe request, it may be identified based on the measurement date and time of each RSSI based on the probe request.

For example, in FIG. 8, when the self-position estimating device 50 transmits a probe request with sequence number "001" at point A, each AP 40 (here, AP40 ₁ to AP40 ₅₎ that detected the probe request The RSSI of the probe request is measured. In this case, for example, the APs 40 ₁ to 40 ₅ may generate RSSI information (for example, a record of point A in FIG. 3) that associates the RSSI with the date and time when the RSSI was measured. Similarly, RSSI information (for example, point B and C in FIG. C record, etc.) is generated at each AP 40 that detects the probe request.

Further, the self-position estimating device 50 estimates its own position (points A, B, C, etc. in FIG. 8) based on the output value of a sensor (for example, a laser sensor, etc.), and calculates its own position on the environmental map. The location information (for example, a two-dimensional coordinate position (x, y) or a three-dimensional coordinate position (x, y, z)) is generated. In this case, the self-position estimating device 50 may generate position-related information (for example, records of points A, B, and C in FIG. 4) that associates the position information with the estimated date and time of the position information.

The learning device 20 may generate learning data that associates the RSSI of each AP 40 within a predetermined area with the location information based on the measurement date and time of the RSSI and the estimated date and time of the location information. Specifically, the learning device 20 associates the location information with RSSI that is associated with one or more measurement dates and times existing within a predetermined time range (for example, 1 second) determined based on the estimated date and time of the location information. Learning data to be associated may be generated. For example, by matching the date and time (measurement date and time) in the record of point A in FIG. 3 with the date and time (estimated date and time) in the record of point A in FIG. 19 minutes 20.254 seconds" (that is, the estimated date and time of the location information of point A in FIG. 4), the RSSI of AP40 ₁ to 40 ₁₁ within a predetermined area may be associated with the location information of point A. .

As mentioned above, the RSSI measured by one or more APs 40 that have detected the same probe request (for example, the three RSSIs corresponding to points A, B, and C in FIG. 3) correspond to the sequence number of the probe request. It may be grouped based on the measurement date and time of the RSSI. Further, the association between the grouped RSSI and the location information may be determined based on the date and time associated with the RSSI and the location information.

In this way, by moving the self-position estimating device 50 within the predetermined area, the RSSI of each AP 40 within the predetermined area and the position information indicating the position of the self-position estimating device 50 estimated by the self-position estimating function can be obtained. Corresponding learning data can be efficiently generated. In addition, the learning data includes all APs 40 in the predetermined area, regardless of whether a signal (for example, a probe request or a data signal) from the self-location device 50 is detected in each AP 40 in the predetermined area. RSSI is included, so the performance of the position estimation model generated based on the learning data can be improved.

<Position estimation operation>
FIG. 9 is a diagram illustrating an example of the position estimation operation according to this embodiment. As shown in FIG. 9, in step S201, the position estimating device 10 acquires the RSSI (also referred to as the RSSI of the user terminal 30) measured using a signal from the user terminal 30 (for example, a probe request) from each AP 40. do. Note that the details of step S201 are the same as step S101 in FIG.

In step S202, the position estimation device 10 inputs the RSSI of the user terminal 30 acquired from each AP 40 in step S201 into the position estimation model, and calculates the position of the user terminal 30 within the predetermined area as an output of the position estimation model. It may be estimated. Furthermore, the position estimating device 10 generates estimated position information indicating the estimated position.

In step S203, the position estimating device 10 may generate information indicating the flow line of the user terminal 30 within a predetermined area (flow line information) based on the estimated position information generated in step S202.

FIG. 10 is a diagram illustrating an example of generation of flow line information of the user terminal 30 according to the present embodiment. In FIG. 10, it is assumed that the user terminal 30 transmits a probe request at a predetermined period in a predetermined area (for example, a shopping mall). Note that, of course, the above data signal or the like may be used instead of the probe request.

For example, when the user terminal 30 transmits a probe request at point A' in FIG. 10, the RSSI of the probe request is measured at each of APs 40 ₁ to 40 ₅ that have detected the probe request. On the other hand, a probe request from the user terminal 30 is not detected in each of the APs 40 ₁ to 40 ₁₁ . The position estimation device 10 inputs the RSSI at AP40 ₁ to AP40 ₁₁ within the time range in which the same probe request is estimated to have been measured into the position estimation model, and calculates the position of point A' as the output of the position estimation model. It is also possible to generate location information indicating the location. Similarly, the position estimating device 10 is based on the RSSI of all APs 40 within a predetermined area within a time range in which probe requests transmitted at different points B' and C' are estimated to have been measured as the user terminal 30 moves. Then, position information indicating the respective positions of the points B' and C' may be generated.

The user terminal 30 such as a smartphone is generally equipped with a wireless LAN communication function. Therefore, by inputting the RSSI of the probe request transmitted from the user terminal 30 into the position estimation model and obtaining the position information of the transmission point of the probe request, the flow line of the user terminal 30 can be suppressed while suppressing the introduction cost. can be understood.

As described above, according to the position estimation system according to the present embodiment, the position of the user terminal 30 within a predetermined area is estimated by inputting the RSSI in the wireless LAN into the position estimation model. The estimation accuracy can be improved.

(Other embodiments)
Note that in the above embodiment, the RSSI is measured by the AP 40 based on a signal from the terminal (position estimating device 10 or user terminal 30), but the present invention is not limited to this. RSSI may be measured at the terminal based on the signal from AP 40. In this case, the position estimating device 10 may acquire the RSSI information from the terminal.

Further, in the above embodiment, the RSSI measured by one or more APs 40 based on the same signal from the self-position estimating device 50 is input, and the position information of the self-position estimating device 50 within the predetermined area is output, Although the learning process of the position estimation model is performed, the present invention is not limited thereto. The learning process of the position estimation model takes the RSSI and other parameters as input, and outputs the position information (for example, 2-dimensional coordinate position (x, y) or 3-dimensional coordinate position (x, y, z)). It may also be implemented as

For example, if a certain shelf has multiple tiers (lower, middle, upper, etc.) in a predetermined area (for example, a warehouse, etc.), and if the self-position estimating device 50 is configured with an unmanned aerial vehicle such as a drone, three-dimensional coordinates By acquiring the position (x, y, z) using the self-position estimation function, it is possible to specify which shelf on a certain display shelf. The learning process of the position estimation model may be performed using other parameters in addition to the RSSI as input parameters and the three-dimensional coordinate position (x, y, z) as an output parameter. In this case, the flow line of the user terminal 30 in three dimensions can be grasped by inputting the RSSI of the user terminal 30 and the other parameters into the position estimation model to obtain the three-dimensional coordinate position.

The embodiments described above are intended to facilitate understanding of the present invention, and are not intended to be interpreted as limiting the present invention. Each element included in the embodiment, as well as its arrangement, material, conditions, shape, size, etc., are not limited to those illustrated, and can be changed as appropriate. Further, it is possible to partially replace or combine the structures shown in different embodiments.

DESCRIPTION OF SYMBOLS 1... Position estimation system, 10... Position estimation device, 20... Learning device, 30... User terminal, 40... Access point (AP), 50... Self-position estimation device, 11... RSSI information acquisition part 12... Estimation part, 13... Storage unit, 14...Control unit, 21...RSSI information acquisition unit, 22...Location-related information acquisition unit, 23...Generation unit, 24...Storage unit, 25...Learning unit, 10a...CPU, 10b...Storage device, 10c...Communication section, 10d...input section, 10e...output section

Claims (7)

Generated by machine learning using learning data that associates received signal strength information indicating the strength of signals from a predetermined device at one or more access points with position information indicating the position of the predetermined device within a predetermined area. a storage unit that stores a position estimation model to be used;
an acquisition unit that acquires received signal strength information indicating the strength of a signal from a user terminal at the one or more access points;
an estimation unit that inputs the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area;
Equipped with
The received signal strength information and the position in the learning data are determined based on whether the date and time associated with the received signal strength information and the date and time when the position was estimated by the predetermined device are within a predetermined time range. The correspondence with the information is determined,
Location estimation device. Correspondence between the received signal strength information and the location information in the learning data based on whether the terminal identification information associated with the received signal strength information and the terminal identification information associated with the location information are the same. The attachment is determined,
The position estimation device according to claim 1 . The signal from the predetermined device is a probe request from the predetermined device, or a data signal transmitted from the predetermined device after a connection is established between the predetermined device and each access point.
The position estimation device according to claim 1 or claim 2 . The position information is a point of two-dimensional coordinates or three-dimensional coordinates estimated using a sensor installed in the predetermined device,
A position estimating device according to any one of claims 1 to 3 . an output unit that outputs information indicating a flow line of the user terminal based on the estimated position of the user terminal;
The position estimating device according to any one of claims 1 to 4 , further comprising:. Generated by machine learning using learning data that associates received signal strength information indicating the strength of signals from a predetermined device at one or more access points with position information indicating the position of the predetermined device within a predetermined area. A position estimation method using a position estimation device using a position estimation model, the method comprising:
the position estimating device acquiring received signal strength information indicating the strength of a signal from the user terminal at the one or more access points;
the position estimation device inputting the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area;
Equipped with
The received signal strength information and the position in the learning data are determined based on whether the date and time associated with the received signal strength information and the date and time when the position was estimated by the predetermined device are within a predetermined time range. The correspondence with the information is determined,
Location estimation method. Generated by machine learning using learning data that associates received signal strength information indicating the strength of signals from a predetermined device at one or more access points with position information indicating the position of the predetermined device within a predetermined area. a calculation unit equipped with a position estimation device that stores a position estimation model to be
an acquisition unit that acquires received signal strength information indicating the strength of a signal from a user terminal at the one or more access points;
an estimation unit that inputs the acquired received signal strength information into the position estimation model to estimate the position of the user terminal within the predetermined area;
function as
The received signal strength information and the position in the learning data are determined based on whether the date and time associated with the received signal strength information and the date and time when the position was estimated by the predetermined device are within a predetermined time range. The correspondence with the information is determined,
Location estimation program.

Images