JP2016170011A - Positioning device, method for positioning, positioning program, and positioning system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately estimate the position of a terminal indoors.SOLUTION: There is provided a positioning device 1 for measuring the position of an interior terminal 2, which includes receiving means 11 receiving, from the terminal 2, terminal information on the terminal 2, sensor information with at least one of a radio-field intensity and a sound-wave intensity measured by the terminal 2, and positional information on the terminal 2 and model generating means 12 generating a learning model for each terminal information with the sensor information and the positional information as teacher data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to indoor positioning technology.

  As a positioning technique, a high-precision positioning technique used for GPS, quasi-zenith satellites, and the like is realized outdoors. On the other hand, a positioning technique using Wi-Fi, BLE (Bluetooth Low Energy (registered trademark)) or the like has been proposed indoors.

  In addition, statistical machine learning for analyzing complex data has been studied (Non-Patent Document 1).

Takashi Takeuchi and three others, "Composite matrix decomposition under non-negative constraints", Information Processing Society of Japan research report. MPS, mathematical modeling and problem solving research report, Information Processing Society of Japan, 2013-05-16, 2013-MPS-93 (3), 1-6 Start position Start position

  Although a positioning technique using Wi-Fi or BLE has been proposed indoors, a highly accurate positioning technique has not been established as compared with the outdoors.

  In addition, radio waves transmitted from a Wi-Fi access point or a BLE transmitter have different reception accuracy depending on the terminal model and the like, so that sufficient positioning accuracy may not be obtained.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to estimate the position of a terminal with high accuracy indoors.

  In order to solve the above-described problem, a first aspect of the present invention is a positioning device that measures the position of an indoor terminal, and includes at least one of terminal information of the terminal, radio wave intensity and sound wave intensity measured by the terminal. Receiving means for receiving the sensor information including the terminal information from the terminal, model generating means for generating a learning model for each terminal information, using the sensor information and the position information as teacher data, Is provided.

  In the positioning device, the receiving unit receives a positioning request including the terminal information of the terminal and the sensor information, selects a learning model corresponding to the terminal information, and adds the sensor information to the selected learning model. Position estimation means for inputting and estimating the position of the terminal is further provided.

  In the positioning device, the model generation unit receives correct position information for the position information estimated by the position estimation unit from the terminal, and uses the correct position information and the sensor information of the positioning request as teacher data. Input to the selected learning model and update the learning model.

  In the positioning device, the model generation unit updates the learning model by adding the teacher data for a position whose positioning accuracy is equal to or less than a predetermined threshold.

  2nd this invention is the positioning method which measures the position of the indoor terminal which a computer performs, Comprising: The sensor information containing the terminal information of the said terminal, and at least one of the radio field intensity and the sound wave intensity which the said terminal measured, A reception step of receiving the position information of the terminal from the terminal, and a model generation step of generating a learning model for each of the terminal information using the sensor information and the position information as teacher data.

  According to a third aspect of the present invention, there is provided a positioning program for measuring the position of an indoor terminal executed by a computer, the computer comprising at least one of terminal information of the terminal and radio wave intensity and sound wave intensity measured by the terminal. Receiving means for receiving the sensor information including the terminal information from the terminal, and model generating means for generating a learning model for each terminal information using the sensor information and the position information as teacher data, To function as.

  4th this invention is a positioning system which has a terminal and the positioning apparatus which measures the indoor position of the said terminal, Comprising: The said terminal is arrange | positioned indoors, the terminal information collection means which collects the terminal information of the said terminal Sensor information collecting means for measuring and collecting sensor information including at least one of radio wave intensity and sound wave intensity of radio waves transmitted from each transmitter, and position information collecting means for collecting position information of the terminal; Transmission means for transmitting the terminal information, the sensor information, and the position information to the positioning device, and the positioning device includes the terminal information, the sensor information, and the position information. Receiving means for receiving from the terminal; and model generating means for generating a learning model for each terminal information using the sensor information and the position information as teacher data.

  5th this invention is a terminal, Comprising: The terminal information collection means which collects the terminal information of the said terminal, At least one of the radio field intensity and sound wave intensity of the radio wave transmitted from each transmitter arrange | positioned indoors Sensor information collecting means for measuring and collecting sensor information, position information collecting means for collecting position information of the terminal, the terminal information, the sensor information, and the position information Transmitting means for transmitting to a positioning device for measuring the position.

  According to the present invention, the position of the terminal can be estimated with high accuracy indoors.

It is a functional block diagram which shows the structure of the indoor positioning system in embodiment. It is a flowchart which shows the flow of the process in which the positioning server of 1st Embodiment produces | generates and updates a learning model. It is a flowchart which shows the flow of the process in which the positioning server of 1st Embodiment estimates a position. It is a flowchart which shows the flow of the process in which the positioning server of 2nd Embodiment estimates a position. It is a figure which shows an example of the image data displayed on a positioning client in 2nd Embodiment. It is a figure which shows an example of the indoor positioning environment in 3rd Embodiment. It is a flowchart which shows the flow of the determination process whether the positioning server of 3rd Embodiment performs additional learning. It is a figure which shows an example of the multidimensional graph by which sensor information was mapped in 3rd Embodiment.

  Embodiments of the present invention will be described below.

<First Embodiment>
FIG. 1 is a functional block diagram showing the configuration of the indoor position positioning system in the first embodiment. The indoor positioning system includes a positioning server 1 (positioning device) and a positioning client 2 (terminal).

  The positioning server 1 receives sensor information such as radio wave intensity or sound wave intensity measured by the positioning client 2 indoors, learns the relationship between the sensor information and the indoor position by machine learning, and generates a learning model. In addition, the positioning server 1 estimates the position of the positioning client 2 by using the generated learning model and the sensor information received from the positioning client 2 as an input.

  The positioning client 2 receives each radio wave or each sound wave transmitted from various transmitters 31-33 installed indoors, and the ID (identification information) of the transmitter 31-33 from which the radio wave or sound wave is transmitted. The radio wave intensity or the sound wave intensity is collected as sensor information and transmitted to the positioning server 1.

  The positioning client 2 shown in FIG. 1 includes a sensor information collection unit 21, a learning position information collection unit 22, and a terminal information collection unit 23. For the positioning client 2, for example, a mobile terminal such as a smartphone can be used.

  The sensor information collection unit 21 receives a Wi-Fi radio wave, a BLE radio wave, a sound wave, etc., and measures and collects ID indicating the radio wave or sound wave transmission source and the radio wave intensity or the sound wave intensity as sensor information. To do.

  The BLE radio wave is a beacon signal transmitted from each of the plurality of BLE transmitters 31, and the Wi-Fi radio wave is transmitted from a plurality of Wi-Fi transmitters 32 (for example, Wi-Fi access points). Beacon signal. The sound wave is a sound wave (ultrasonic wave) that is transmitted from the plurality of sound wave transmitters 33 and cannot be heard by humans.

  The learning position information collection unit 22 collects and acquires the current position (correct position) of the positioning client 2 indoors as learning position information. The learning position information may be input by a user using the positioning client 2, for example.

  The terminal information collection unit 23 collects and acquires terminal information (model information) of the positioning client 2. The terminal information includes the OS type, OS version, model name, vendor, etc. of the positioning client 2. It is assumed that the terminal information is stored in advance in a storage unit (not shown) of the positioning client 2.

  When the learning model is generated (during learning), the transmission unit 24 includes the sensor information collected by the sensor information collecting unit 21, the learning position information collected by the learning position information collecting unit 22, and the terminal information collecting unit 23. The collected terminal information is transmitted to the positioning server 1. Further, when using the learning model (at the time of positioning request), the transmitting unit 24 sends a positioning request including the sensor information collected by the sensor information collecting unit 21 and the terminal information collected by the terminal information collecting unit 23 to the positioning server. 1 to send.

  The receiving unit 25 receives the position information of the positioning client 2 estimated by the positioning server 1 in response to the positioning request.

  The positioning server 1 shown in FIG. 1 includes a receiving unit 11, a model generating unit 12, a model DB 13, a position estimating unit 14, and a transmitting unit 15.

  The receiving unit 11 receives sensor information, learning position information, and terminal information from the positioning client 2 at the time of learning model generation. The receiving unit 11 receives a positioning request including sensor information and terminal information from the positioning client 2 when using the model (when estimating the position).

  The sensor information includes an ID indicating a transmission source of radio waves and / or sound waves such as Wi-Fi and BLE collected and measured by the positioning client 2, and radio wave intensity or sound wave intensity. The learning position information is the current position of the positioning client 2. The terminal information includes the OS type, OS version, model name, vendor, etc. of the positioning client 2.

  The model generation unit 12 generates and updates a learning model for estimating the position of the positioning client 2 based on the information received from the positioning client 2. That is, the model generation unit 12 generates a learning model for each terminal information using the sensor information and the learning position information as teacher data. The learning model is a model (function) that outputs the estimated position of the positioning client 2 when certain sensor information is input. In the present embodiment, the learning model is generated for each terminal information of the positioning client 2.

  As an example of machine learning for generating a learning model, for example, a multidimensional machine learning technique using “Non-Negative Multiple Matrix Factorization (NMMF)” described in Non-Patent Document 1, “Self-Organizing Incremental Natural Network (NSO)” ) "," Deep Natural Networks (DNN) ", and the like.

  In this embodiment, the radio wave intensity of each radio wave and / or the sound wave intensity of each sound wave measured by the positioning client 2 at each indoor point is combined and analyzed in a multidimensional manner to generate a learning model.

For example, in the sensor information, as shown below, for each radio wave or each sound wave, the transmitter ID and the radio wave intensity or the sound wave intensity are set in pairs.
“(BLE transmitter A: radio wave intensity a), (BLE transmitter B: radio wave intensity b), (BLE transmitter C: radio wave intensity c), (Wi-Fi transmitter A: radio wave intensity d), (Wi− Fi transmitter B: radio wave intensity e), (Wi-Fi transmitter C: radio wave intensity f), (sound wave transmitter A: radio wave intensity g) "
The sensor information includes at least one of the radio wave intensity and the sound wave intensity, and may include only the radio wave intensity or only the sound wave intensity.

  The model DB 13 stores the learning model generated by the model generation unit 12 for each terminal information.

  The position estimation unit 14 uses the learning model stored in the model DB 13 and corresponding to the terminal information received from the positioning client 2 to estimate the position of the positioning client 2 using the sensor information received from the positioning client 2 as an input. . That is, the position estimation unit 14 selects a learning model corresponding to the terminal information from the model DB 13, inputs sensor information to the selected learning model, and estimates the output of the learning model as the position of the terminal.

  In addition, when the learning model corresponding to the received terminal information does not exist in the model DB 13, the position is estimated using a learning model of terminal information similar to the terminal information (for example, terminal information of the same OS and model).

  The transmission unit 15 transmits the position of the positioning client 2 estimated by the position estimation unit 14 to the positioning client 2.

  As the positioning server 1 and the positioning client 2 described above, for example, a general-purpose computer system including a CPU, a memory, and an external storage device such as a hard disk can be used. In this computer system, each function of each device is realized by the CPU executing a predetermined program loaded on the memory. For example, each function of the positioning server 1 and the positioning client 2 is executed by the CPU of the positioning server 1 in the case of the program for the positioning server 1, and by the CPU of the positioning client 2 in the case of the program for the positioning client 2, respectively. It is realized by doing.

  Further, the program for the positioning server 1 and the program for the positioning client 2 can be stored in a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, a DVD-ROM, or distributed via a network. You can also

  Next, the processing of this embodiment will be described.

  FIG. 2 is a flowchart showing a flow of processing in which the positioning server 1 generates and updates a learning model.

  The receiving unit 11 of the positioning server 1 receives sensor information, learning position information, and terminal information from the positioning client 2 (step S11).

  The model generation unit 12 determines whether a learning model corresponding to the received terminal information (a learning model corresponding to the positioning client 2) exists in the model DB 13 (step S12).

  When there is a learning model corresponding to the received terminal information (step S12: YES), the model generation unit 12 selects the learning model (step S13).

  When there is no learning model corresponding to the received terminal information (step S12: NO), the model generation unit 12 generates an initial learning model corresponding to the terminal information (step S14). Here, the model generation unit 12 uses the learning model of the terminal information that is most similar to the received terminal information among the learning models that exist in the model DB 13, and sets it as the initial learning model. For example, it is conceivable to use a learning model of terminal information having the same model and OS and the closest OS version as a learning model of the most similar terminal information.

  The model generation unit 12 inputs the sensor information and the learning position information received from the positioning client 2 as teacher data to the learning model selected and generated in step S13 or S14. A learning model is generated and updated so that position information is output (becomes correct) (step S15).

  FIG. 3 is a flowchart showing a flow of processing in which the positioning server 1 estimates the position of the positioning client 2.

  The receiving unit 11 of the positioning server 1 receives a positioning request including sensor information and terminal information from the positioning client 2 (step S21).

  The position estimation unit 14 determines whether or not a learning model corresponding to the received terminal information exists in the model DB 13 (step S22).

  When the learning model corresponding to the received terminal information exists (step S22: YES), the position estimation unit 14 selects the learning model (step S23).

  When there is no learning model corresponding to the received terminal information (step S22: NO), the position estimation unit 14 selects a general-purpose learning model (step S24). Here, the position estimation unit 14 may select, as a general-purpose learning model, a learning model of terminal information that is most similar to the received terminal information among learning models existing in the model DB 13.

  The position estimation unit 14 inputs the sensor information received from the positioning client 2 to the learning model selected in step S23 or S24, and estimates the position of the positioning client 2 (step S25).

  The transmission unit 15 transmits the position information of the position estimated by the position estimation unit 14 to the positioning client 2 (step S26).

<Second Embodiment>
In the second embodiment, teacher data of a learning model is collected in a CGM (Consumer Generated Media) manner.

  In the first embodiment described above, the learning position information is collected only when the learning model is generated, and the learning model is generated and updated. On the other hand, in this embodiment, even when the learning model is used (position estimation), correct position information input by the user of the positioning client 2 is collected and the learning model is updated.

  Since the configuration of this embodiment is the same as that of the first embodiment (FIG. 1), description thereof is omitted here.

  FIG. 4 is a flowchart showing a flow of processing in which the positioning server 1 estimates the position of the positioning client 2 in the present embodiment.

  Steps S31 to S35 shown in FIG. 4 are the same as steps S21 to S25 of the process shown in FIG.

  In step S <b> 36, the transmission unit 15 transmits the position information of the position estimated by the position estimation unit 14 to the positioning client 2. In the present embodiment, the transmission unit 15 transmits image data indicating the positioning target indoors to the positioning client 2 together with the estimated position information. The positioning client 2 displays the transmitted image data.

  FIG. 5A is an example of image data displayed on the display of the positioning client 2. In the illustrated image data, an estimated position “×” estimated by the positioning server 1 is set. The image data is generated by the position estimation unit 14.

  When the estimated position “×” of the image data is different from the correct position recognized by the user of the positioning client 2, as shown in FIG. 5B, the correct position “◯” recognized by the user is displayed on the image data. Enter and click the Learn button (not shown). The learning position information collection unit 22 of the positioning client 2 transmits the position information of the correct position input by the user to the positioning server 1 via the transmission unit 24.

  The model generation unit 12 of the positioning server 1 receives the position information of the correct position with respect to the estimated position information via the reception unit 11 (step S37).

  The model generation unit 12 inputs the received position information of the correct position and the sensor information of the positioning request received in S31 as teacher data to the learning model selected or generated in step S33 or S34, and updates the learning model (Step S38). That is, the model generation unit 12 updates the learning model so that the position information of the correct position is output when sensor information is input.

  As described above, in this embodiment, the learning model can be refined by collecting information on the correct position recognized by the user who has made the positioning request and updating the learning model as teacher data.

<Third Embodiment>
The third embodiment is an embodiment in which additional learning is performed on a point with poor positioning accuracy.

  As shown in FIG. 6, objects such as walls, desks, and foliage plants are arranged indoors, and the positioning environment is not uniform. Therefore, there are points that require a lot of teacher data (points with poor positioning accuracy) and points that have a small amount of teacher data (points with good positioning accuracy). In the illustrated example, among the learning points, points in the vicinity of an object such as a wall are indicated as points with poor positioning accuracy.

  In the present embodiment, the teacher data is further collected for a point with low accuracy and input to the learning model for additional learning to improve the accuracy of the learning model at the point.

  Since the configuration of this embodiment is the same as that of the first embodiment (FIG. 1), description thereof is omitted here.

  In the present embodiment, after the learning model is generated, the accuracy of the estimated position is calculated at each point. For example, at each learning point x shown in FIG. 6, the positioning server 1 calculates the accuracy based on the difference between the position estimated using the learning model and the position where the positioning client 2 is actually located (learning point x). However, if the accuracy is less than or equal to the threshold value, additional learning is performed.

  FIG. 7 is a flowchart illustrating a flow of a determination process for determining whether or not the positioning server 1 according to the present embodiment performs additional learning at each learning point x.

  Steps S41 to S45 shown in FIG. 7 are the same as steps S21 to S25 of the process shown in FIG.

  In step S46, the position estimation unit 14 calculates a difference (error) between the estimated position and the learning point where the positioning client 2 is actually located, and calculates accuracy using the difference. Note that the smaller the difference, the higher the accuracy, and the larger the difference, the lower the accuracy.

  Then, the position estimation unit 14 determines whether or not the calculated accuracy is equal to or less than a predetermined threshold (step S47), and performs additional learning for points that are equal to or less than the threshold (step S48).

  That is, for a point that is equal to or less than a predetermined threshold, the model generation unit 12 inputs additional teacher data to the learning model by repeatedly performing the model creation process illustrated in FIG. Update.

  FIG. 8A is a diagram in which the sensor information of the teacher data at the point P with good positioning accuracy is mapped to a multidimensional graph. FIG. 8A shows that the cluster is relatively clear even if the teacher data is small, and thus the positioning accuracy of the learning model is high.

  On the other hand, FIG. 8B is a diagram in which the sensor information of the teacher data after additional learning is mapped to a multidimensional graph at the point Q where the positioning accuracy is poor. In FIG. 8B, the number of teacher data increases due to additional learning, so that the boundaries of the clusters become clear and the positioning accuracy of the learning model is improved.

  Note that “◯” shown in FIG. 8 is teacher data in a certain cluster (BLE-A, BLE-B, Wi-Fi-α in the illustrated example), and “×” is a teacher data in another cluster. Data is shown. A cluster is generated for each combination of radio waves and / or sound waves transmitted from at least one transmitter having a predetermined ID.

  In the first to third embodiments described above, the positioning server 1 includes the terminal information of the positioning client 2, sensor information including at least one of the radio wave intensity and the sound wave intensity measured by the positioning client 2, and the positioning client 2. Is received from the positioning client 2, and a learning model is generated for each terminal information using the sensor information and the position information as teacher data. In this embodiment, since a learning model is generated for each terminal information that affects positioning, such as the OS type and model of the positioning client 2, positioning with higher accuracy can be performed even indoors.

  In the second embodiment, even when the learning model is used (position estimation), correct position information input by the user who has made the positioning request is collected and the learning model is updated. Thereby, even when the learning model is used, the learning model can be updated and further refined.

  In the third embodiment, for points with poor positioning accuracy, teacher data is added, additional learning is performed, and the learning model is updated. Thereby, even if it is a point where positioning accuracy is originally bad, positioning accuracy can be improved.

  In addition, this invention is not limited to said embodiment, Many deformation | transformation are possible within the range of the summary.

1: positioning server 11: receiving unit 12: model generating unit 13: model DB
14: Position estimation unit 15: Transmission unit 2: Positioning client 21: Sensor information collection unit 22: Learning position information collection unit 23: Terminal information collection unit 24: Transmission unit 25: Reception unit

In order to solve the above-described problem, a first aspect of the present invention is a positioning device that measures the position of an indoor terminal, and includes at least one of terminal information of the terminal, radio wave intensity and sound wave intensity measured by the terminal. Receiving means for receiving the sensor information including the terminal information from the terminal, model generating means for generating a learning model for each terminal information, using the sensor information and the position information as teacher data, The receiving means receives a positioning request including terminal information of the terminal and the sensor information, selects a learning model corresponding to the terminal information, and inputs the sensor information to the selected learning model. And a position estimating means for estimating the position of the terminal.

2nd this invention is the positioning method which measures the position of the indoor terminal which a computer performs, Comprising: The sensor information containing at least one of the terminal information of the said terminal, and the electromagnetic wave intensity and sound wave intensity which the said terminal measured When, the position information of the terminal, a receiving step of receiving from the terminal, the sensor information and the position information as teacher data, have rows in the model generating step and the generating a learning model for each of the terminal information, A positioning request including the terminal information of the terminal and the sensor information is received, a learning model corresponding to the terminal information is selected, the sensor information is input to the selected learning model, and the position of the terminal is estimated A position estimation step is further performed.

3rd this invention is a positioning program which measures the position of the indoor terminal which a computer performs, Comprising: The said computer is the terminal information of the said terminal, At least 1 of the radio wave intensity | strength and sound wave intensity which the said terminal measured. Receiving means for receiving sensor information including two and position information of the terminal from the terminal, and model generating means for generating a learning model for each terminal information using the sensor information and the position information as teacher data The receiving means receives a positioning request including the terminal information of the terminal and the sensor information, selects a learning model corresponding to the terminal information, and adds the sensor information to the selected learning model. It is further made to function as position estimation means for inputting and estimating the position of the terminal.

4th this invention is a positioning system which has a terminal and the positioning apparatus which measures the indoor position of the said terminal, Comprising: The said terminal is arrange | positioned indoors, the terminal information collection means which collects the terminal information of the said terminal Sensor information collecting means for measuring and collecting sensor information including at least one of radio wave intensity and sound wave intensity of radio waves transmitted from each transmitter, and position information collecting means for collecting position information of the terminal; Transmission means for transmitting the terminal information, the sensor information, and the position information to the positioning device, and the positioning device includes the terminal information, the sensor information, and the position information. receiving means for receiving from the terminal, the sensor information and the position information as teacher data, and a model generation means for generating a learning model for each of the terminal information, said receiving means, said A position for receiving a positioning request including terminal information at the end and the sensor information, selecting a learning model corresponding to the terminal information, inputting the sensor information to the selected learning model, and estimating a position of the terminal An estimation means is further provided .

  In the above positioning apparatus, the position estimation unit selects a learning model of terminal information that is most similar to the terminal information when there is no learning model corresponding to the terminal information.

As an example of machine learning that generates a learning model, for example, a multi-dimensional machine learning technique using “Non-Negative Multiple Matrix Factorization (NMMF)” described in Non-Patent Document 1, “Self-Organizing Incremental Natural Network (NSO)” ) "," Deep Neural Networks (DNN) ", and the like.

Claims (8)

A positioning device that measures the position of an indoor terminal,
Receiving means for receiving terminal information of the terminal, sensor information including at least one of radio wave intensity and sound wave intensity measured by the terminal, and position information of the terminal;
A positioning device comprising: model generation means for generating a learning model for each of the terminal information using the sensor information and the position information as teacher data. The positioning device according to claim 1,
The receiving means receives a positioning request including terminal information of the terminal and the sensor information,
A positioning apparatus, further comprising: a position estimation unit that selects a learning model corresponding to the terminal information, inputs the sensor information to the selected learning model, and estimates a position of the terminal. The positioning device according to claim 2,
The model generation means receives correct position information for the position information estimated by the position estimation means from the terminal, and the selected learning model uses the correct position information and the sensor information of the positioning request as teacher data. And the learning model is updated. The positioning device according to any one of claims 1 to 3,
The model generation means updates the learning model by adding the teacher data for a position whose positioning accuracy is a predetermined threshold value or less. A positioning method for measuring the position of an indoor terminal performed by a computer,
Receiving the terminal information of the terminal, sensor information including at least one of radio wave intensity and sound wave intensity measured by the terminal, and position information of the terminal from the terminal;
A model generation step of generating a learning model for each of the terminal information using the sensor information and the position information as teacher data. A positioning program for measuring the position of an indoor terminal executed by a computer,
The computer,
Receiving means for receiving terminal information of the terminal, sensor information including at least one of radio wave intensity and sound wave intensity measured by the terminal, and position information of the terminal; and
A positioning program that functions as model generation means for generating a learning model for each terminal information using the sensor information and the position information as teacher data. A positioning system having a terminal and a positioning device that measures the indoor position of the terminal,
The terminal
Terminal information collecting means for collecting terminal information of the terminal;
Sensor information collecting means for measuring and collecting sensor information including at least one of radio wave intensity and sound wave intensity of radio waves transmitted from each transmitter disposed indoors;
Position information collecting means for collecting position information of the terminal;
Transmission means for transmitting the terminal information, the sensor information, and the position information to the positioning device;
The positioning device is
Receiving means for receiving the terminal information, the sensor information, and the position information from the terminal;
And a model generation unit that generates a learning model for each of the terminal information using the sensor information and the position information as teacher data. A terminal,
Terminal information collecting means for collecting terminal information of the terminal;
Sensor information collecting means for measuring and collecting sensor information including at least one of radio wave intensity and sound wave intensity of radio waves transmitted from each transmitter disposed indoors;
Position information collecting means for collecting position information of the terminal;
A terminal comprising: transmission means for transmitting the terminal information, the sensor information, and the position information to a positioning device that measures an indoor position of the terminal.

Images