Learning model creation device, learning model creation system, and program

Abstract

To easily acquire learning data used to create a learning model to determine the position of a wireless tag.SOLUTION: The learning model creation device includes a first creation unit and a second creation unit. The first creation unit creates virtual tag data regarding a virtual tag on the basis of tag data regarding a wireless tag. The second creation unit creates a learning model to determine the position of a determination target wireless tag on the basis of tag data regarding the determination target wireless tag by a machine learning based on learning data including virtual tag data regarding a plurality of virtual wireless tags created by the first creation unit.SELECTED DRAWING: Figure 15

Description

[Technical Field]

[0001] Embodiments of the present invention relate to a learning model creation apparatus, a learning model creation system, and a program. [Background Art]

[0002] There is an apparatus that uses a learning model to determine whether a wireless tag is a reading target based on tag data such as the phase of the wireless tag. Such an apparatus acquires tag data in advance at a plurality of coordinates where a wireless tag is expected to be placed, and creates a learning model that determines whether or not a target is a reading target based on learning data including the acquired tag data. The apparatus can improve the determination accuracy using the learning model by acquiring tag data in advance at all coordinates where a wireless tag is expected to be placed.

[0003] However, in order to create a learning model with high determination accuracy, the apparatus must acquire tag data at all coordinates as described above. Therefore, the apparatus requires a lot of time to acquire learning data used for creating a learning model with high determination accuracy. [Prior Art Literature] [Patent Literature]

[0004] [Patent Literature 1] Japanese Unexamined Patent Publication No. 2021-047723 [Summary of the Invention] [Problem to be Solved by the Invention]

[0005] The problem to be solved by the embodiments of the present invention is to provide a technique that enables easy acquisition of learning data used for creating a learning model for determining the position of a wireless tag. [Means for Solving the Problem]

[0006] The learning model creation apparatus of the embodiment comprises a first creation unit and a second creation unit. The first creation unit creates virtual tag data for virtual tags based on tag data for wireless tags. The second creation unit creates a learning model for determining the location of a target wireless tag based on tag data for the target wireless tag by machine learning based on learning data including virtual tag data for a plurality of virtual wireless tags created by the first creation unit. [Brief explanation of the drawing]

[0007] [Figure 1] Figure 1 is a block diagram showing an example of the configuration of a learning model creation system according to an embodiment. [Figure 2] Figure 2 is a top view illustrating the first determination range and the second determination range according to the embodiment. [Figure 3] Figure 3 is a top view illustrating the movement of the antenna according to this embodiment. [Figure 4] Figure 4 is a top view illustrating the movement of the antenna according to the embodiment. [Figure 5] Figure 5 is a side view illustrating the movement of the antenna according to this embodiment. [Figure 6] Figure 6 is a diagram illustrating the creation of virtual tag data based on tag data according to the embodiment. [Figure 7] Figure 7 is a graph showing tag data related to location PA tags according to the embodiment. [Figure 8] Figure 8 is a graph showing tag data related to the location PB tag according to the embodiment. [Figure 9] Figure 9 is a graph showing virtual tag data related to the virtual tag of location PB according to the embodiment. [Figure 10] Figure 10 is a diagram showing the relationship between tag data for location PA and virtual tag data for location PB according to the embodiment. [Figure 11] Figure 11 is a flowchart showing an example of processing by the processor of the reading device according to the embodiment. [Figure 12] FIG. 12 is a flowchart illustrating an example of processing performed by a processor of the learning model creation apparatus according to the embodiment. [Figure 13] FIG. 13 is a flowchart illustrating an example of data synthesis processing performed by a processor of the learning model creation apparatus according to the embodiment. [Figure 14] FIG. 14 is a flowchart illustrating an example of data shift processing performed by a processor of the learning model creation apparatus according to the embodiment. [Figure 15] FIG. 15 is a diagram for explaining creation of virtual tag data performed by a processor of the learning model creation apparatus according to the embodiment. [Figure 16] FIG. 16 is a top view for explaining a first modification of the reading apparatus according to the embodiment. [Figure 17] FIG. 17 is a side view for explaining a second modification of the reading apparatus according to the embodiment. [Figure 18] FIG. 18 is a top view for explaining the second modification of the reading apparatus according to the embodiment. [Figure 19] FIG. 19 is a top view for explaining the arrangement of a plurality of tags. [Figure 20] FIG. 20 is a graph showing tag data. [Figure 21] FIG. 21 is a plurality of graphs showing tag data. [Figure 22] FIG. 22 is a graph showing tag data respectively for two tags. MODE FOR CARRYING OUT THE INVENTION

[0008] (Embodiment) Hereinafter, embodiments will be described with reference to the drawings. Note that in each drawing used for describing the following embodiments, the scale of each part may be appropriately changed. Further, in each drawing used for describing the following embodiments, some configurations may be omitted for the sake of explanation.

[0009] (Configuration Example) FIG. 1 is a block diagram showing an example configuration of a learning model creation system S. The learning model creation system S includes a reading device 1 and a learning model creation device 2.

[0010] The reading device 1 is a device that performs wireless communication with a tag 3 and reads the tag 3. Reading the tag 3 includes acquiring information stored in the tag 3 and acquiring tag data related to the tag 3. Acquiring tag data includes the meaning of detecting tag data or measuring tag data. The reading device 1 can be applied to inspection in warehouses, and may also be applied in stores; application examples of the reading device 1 are not limited thereto. A configuration example of the reading device 1 will be described later.

[0011] Tag data is data acquired in time series based on radio waves of the tag 3 received by the reading device 1. The radio waves of the tag 3 are radio waves transmitted from the tag 3, and may also be referred to as radio waves from the tag 3. The tag data includes at least one of phase data, Doppler frequency data, and RSSI (Received Signal Strength Indicator) data. The phase data is data indicating the phase of the radio wave of the tag 3 received by the reading device 1. The Doppler frequency data is data indicating the frequency of the radio wave of the tag 3 received by the reading device 1. The RSSI data is data indicating the RSSI of the radio wave of the tag 3 received by the reading device 1. RSSI indicates reception strength. Reception strength is also referred to as radio wave reception strength or received signal strength.

[0012] The learning model creation device 2 is a device that creates a learning model for determining the position of a determination target wireless tag based on tag data related to the determination target wireless tag. The learning model is also referred to as an inference engine. The learning model will be described later. The determination target wireless tag is an entity wireless tag configured in the same manner as the tag 3 described later. The expression "creation" includes not only the aspect of new creation but also the aspect of updating. A configuration example of the learning model creation device 2 will be described later.

[0013] Determining the position of a target wireless tag includes determining the positional relationship of the target wireless tag with respect to a first determination range. For example, the positional relationship of the target wireless tag with respect to the first determination range is that the target wireless tag is within the first determination range or within the second determination range. The first and second determination ranges are distinct ranges that do not overlap with each other. For example, the first and second determination ranges are three-dimensional ranges. The second determination range is the range outside the first determination range. The second determination range is described as a range that is not adjacent to the first determination range, but it may also be a range that is adjacent to the first determination range. Range includes the meaning of area. The first determination range is an example of a predetermined range. Examples of the first and second determination ranges will be described later. The target wireless tag may be a wireless tag within the first determination range or a wireless tag within the second determination range.

[0014] Determining the positional relationship of the target radio tag with respect to the first determination range includes determining whether the target radio tag is within the first determination range or the second determination range. Determining whether the target radio tag is within the first determination range or the second determination range includes determining whether the target radio tag is within the first determination range or the second determination range. Being within the first determination range includes the target radio tag being located within the first determination range. Being within the first determination range may also include considering the target radio tag to be within the first determination range. Being within the second determination range includes the target radio tag being located within the second determination range. Being within the second determination range may also include considering the target radio tag to be within the second determination range.

[0015] Tag 3 is a wireless tag with a physical entity used to create training data. Tag 3 may be located within the first judgment range or within the second judgment range.

[0016] Tag 3 is an IC tag including an IC chip and an antenna. Tag 3 is typically an RFID (Radio Frequency Identification) tag. Tag 3 may be any other type of IC tag. Tag 3 is a passive wireless tag that operates using radio waves transmitted from antenna 16 as an energy source. Tag 3 transmits a signal containing information stored on its IC chip via the antenna by performing backscatter modulation on an unmodulated signal. The information stored on Tag 3 may include uniquely identifiable identification information. Hereinafter, the identification information stored on Tag 3 may be abbreviated as "identification information". Identification information is an example of information about Tag 3. Hereinafter, the identification information stored on Tag 3 will also be referred to as the identification information of Tag 3. Figure 1 shows one Tag 3, but the reader 1 may wirelessly communicate with multiple Tag 3s and read multiple Tag 3s. In this case, multiple Tag 3s are used to create training data.

[0017] An example configuration of the reading device 1 will be described below. The reader device 1 includes a processor 11, a ROM (Read-Only Memory) 12, a RAM (Random-Access Memory) 13, a connection interface 14, a reader / writer 15, an antenna 16, a drive unit 17, and a storage device 18. Each part of the reader device 1 is connected by a bus 19, etc.

[0018] The processor 11 corresponds to the central part of the computer that performs calculations and control and other processing necessary for the operation of the reader 1. The processor 11 includes one or more processors. For example, the processor is a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), but is not limited to these. The processor may be composed of various circuits. The processor 11 loads a program pre-stored in the ROM 12 or storage device 18 into the RAM 13. The program is a program that enables the processor 11 to implement each part described later and to make each part executable by the processor 11. The processor 11 enables the processing of each part by executing the program loaded into the RAM 13. The processor 11 is an example of a processing circuit.

[0019] ROM12 corresponds to the main memory of the computer, with the processor 11 at its core. ROM12 is a non-volatile memory used exclusively for reading data. ROM12 stores the above-mentioned program. ROM12 also stores data or various setting values ​​used by the processor 11 in performing various processes.

[0020] RAM13 corresponds to the main memory of a computer centered around processor 11. RAM13 is memory used for reading and writing data. RAM13 is a work area that stores data temporarily used by processor 11 when performing various processes.

[0021] The connection interface 14 is an interface for the reading device 1 to communicate with the learning model creation device 2. The connection interface 14 may include an interface for wired connection or an interface for wireless connection.

[0022] The reader / writer 15 includes a communication processing circuit that handles communication between the tag 3 and the antenna 16 and reads the tag 3. The reader / writer 15 converts the digital signal to an analog signal. The reader / writer 15 outputs the analog signal to the antenna 16. The reader / writer 15 receives the analog signal from the antenna 16 as input. The reader / writer 15 converts the analog signal input from the antenna 16 into a digital signal. The reader / writer 15 acquires information based on the radio waves of the tag 3 received by the antenna 16. For example, the reader / writer 15 acquires the identification information of the tag 3 from the digital signal using known techniques. The reader / writer 15 acquires tag data in time series based on the radio waves of the tag 3 received by the antenna 16. For example, the reader / writer 15 can acquire phase data in time series from the digital signal using known techniques. The reader / writer 15 can acquire Doppler frequency data in time series from the digital signal using known techniques. The reader / writer 15 can acquire RSSI data in time series from a digital signal using known technology. The reader / writer 15 is an example of an acquisition unit that acquires tag data related to tag 3 at multiple locations on the antenna 16. When the reader / writer 15 includes the antenna 16, the communication processing circuit is an example of an acquisition unit that acquires tag data related to tag 3 at multiple locations on the antenna 16.

[0023] Antenna 16 communicates with tag 3.

[0024] The drive unit 17 moves the antenna 16. For example, the drive unit 17 includes a stage on which the antenna 16 is placed and a motor that drives the stage, but the drive unit 17 may also include other components. Here, the drive unit 17 moves the antenna 16 in a circular direction about a rotation axis along the vertical direction in the horizontal plane. Moving the antenna 16 includes changing the position of the antenna 16. Here, the position of the antenna 16 is a value represented by the rotation angle of the antenna 16. Hereafter, the rotation angle of the antenna 16 may be abbreviated as rotation angle. The rotation angle is a value corresponding to the amount of movement along the direction of movement of the antenna 16. The drive unit 17 moves the antenna 16 from the starting position to the ending position. Here, the starting position is described as the position where the rotation angle is 0°. The position where the rotation angle is 0° can be set as appropriate. The ending position is described as the position where the rotation angle of the antenna 16 is 360°. Therefore, the drive unit 17 moves the antenna 16 so that it makes one full rotation in the circumferential direction around the axis of rotation. In this example, the range of movement of the antenna 16 is from a rotation angle of 0° to 360°. Note that the end position is not limited to a rotation angle of 360°. The end position may be any position with a rotation angle less than 360°. The position of the antenna 16 is an example of the relative position of the antenna 16 with respect to the tag 3. Moving the antenna 16 is an example of changing the relative position of the antenna 16 with respect to the tag 3. The range of movement of the antenna 16 is an example of the range of movement of the relative position of the antenna 16 with respect to the tag 3.

[0025] Although Figure 1 shows one antenna 16, the reading device 1 may use multiple antennas 16 to read the tag 3. In this case, the drive unit 17 moves each of the antennas 16. The movement range of each of the multiple antennas 16 may be different.

[0026] The storage device 18 is a device composed of non-volatile memory for storing data and programs. The storage device 18 is composed of, but is not limited to, an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage device 18 is an example of a storage unit.

[0027] The storage device 18 includes a measurement data storage area 181. The measurement data storage area 181 stores the measurement data. The measurement data includes a tag dataset for each tag 3 read by the reader / writer 15. The tag dataset includes tag data related to tag 3 within the movement range of the antenna 16. The tag data related to tag 3 within the movement range of the antenna 16 is tag data related to tag 3 at multiple positions of the antenna 16 included within the movement range of the antenna 16. The tag data is acquired by the reader / writer 15 in accordance with the movement of the antenna 16 along the circumferential direction around the rotation axis by the drive unit 17. The tag data is associated with the position data of the antenna 16. The position data of the antenna 16 is data indicating the position of the antenna 16. Hereinafter, the tag data related to tag 3 within the movement range of the antenna 16 will also be referred to as tag data within the movement range of the antenna 16. The tag dataset is an example of tag data within the movement range of the antenna 16.

[0028] The multiple positions of the antenna 16 may include positions at regular intervals between the start and end positions within the range of movement of the antenna 16. The value of the regular interval can be set as appropriate. Depending on the tag 3, the reader / writer 15 may acquire tag data for all positions at regular intervals between the start and end positions. Depending on the tag 3, the reader / writer 15 may not acquire tag data for some of the positions at regular intervals between the start and end positions.

[0029] The tag dataset may include phase data for tag 3 within the range of movement of antenna 16. The phase value changes as the position of antenna 16 changes. This is because the distance between antenna 16 and tag 3 changes as antenna 16 moves. Since the phase value depends on the distance between antenna 16 and tag 3, the distribution of phase data differs depending on the position of tag 3.

[0030] The tag dataset may include Doppler frequency data for tag 3 within the range of movement of antenna 16. The Doppler frequency value changes as the position of antenna 16 changes. This is because the Doppler frequency value differs when antenna 16 approaches tag 3 and when antenna 16 moves away from tag 3. The distribution of Doppler frequency data differs depending on the position of tag 3.

[0031] The tag dataset may include RSSI data for tag 3 within the range of movement of antenna 16. The RSSI value changes as the position of antenna 16 changes. This is because the distance between antenna 16 and tag 3 changes as antenna 16 moves. Since the RSSI value depends on the distance between antenna 16 and tag 3, the distribution of RSSI data differs depending on the position of tag 3.

[0032] The measurement data includes the position of each tag 3 read by the reader / writer 15. For example, the position of tag 3 is the coordinates indicating the location of tag 3. The position of tag 3 may be entered by the user. The measurement data can be updated.

[0033] Although an example has been described in which the memory device 18 stores measurement data, the explanation is not limited to this. The RAM 13 may also store measurement data. In this case, the RAM 13 is an example of a memory unit.

[0034] Bus 19 includes a control bus, an address bus, and a data bus, etc. Bus 19 transmits signals exchanged between the various parts of the reader 1.

[0035] The hardware configuration of the reader 1 is not limited to the configuration described above. The reader 1 may be modified or have its components omitted or changed as appropriate, and new components added as needed.

[0036] The processor 11 implements the read processing unit 111 and the communication processing unit 112. Each part implemented by the processor 11 can also be called a function. Each part implemented by the processor 11 can also be said to be implemented by a control unit including the processor 11, ROM 12, and RAM 13.

[0037] The reading processing unit 111 processes the reading of tag 3. The communication processing unit 112 processes communication between the reading device 1 and the learning model creation device 2.

[0038] This section describes an example configuration of the learning model creation device 2. The learning model creation device 2 includes a processor 21, ROM 22, RAM 23, memory device 24, and connection interface 25. The various parts included in the learning model creation device 2 are connected by a bus 26, etc.

[0039] The processor 21 corresponds to the central part of the computer that performs calculations and control processes necessary for the operation of the learning model creation device 2. The processor 21 may be configured in the same way as the processor 11. The processor 21 loads a program pre-stored in the ROM 22 or memory device 24 into the RAM 23. The program is a program that enables the processor 21 to implement each part described later and to make each part executable. The processor 21 enables the processing of each part by executing the program loaded into the RAM 23. The processor 21 is an example of a processing circuit.

[0040] ROM22 corresponds to the main memory of a computer centered around processor 21. ROM22 may be configured similarly to ROM12.

[0041] RAM23 corresponds to the main memory of a computer centered around processor 21. RAM23 may be configured similarly to RAM13.

[0042] The storage device 24 is a device composed of non-volatile memory for storing data and programs, etc. The storage device 24 may be configured in the same way as the storage device 18. The storage device 24 is an example of a storage unit.

[0043] The memory device 24 includes a learning data storage area 241. The training data storage area 241 stores the training data. The training data includes virtual tag data for multiple virtual tags. A virtual tag is a virtual radio tag that does not have a physical form. The virtual tag data for a virtual tag is data created based on the tag data for tag 3. The virtual tag data includes at least one of the following: virtual phase data, virtual Doppler frequency data, and virtual RSSI data. The virtual phase data is data created based on the phase data for tag 3. The virtual Doppler frequency data is data created based on the Doppler frequency data for tag 3. The virtual RSSI is data created based on the RSSI data for tag 3.

[0044] The first creation unit 212 creates virtual tag data for virtual tags based on the tag data for tag 3. For example, the first creation unit 212 creates virtual tag data for virtual tags in a second scope based on the tag data for tag 3 in a first scope. Hereinafter, the tag data for tag 3 in the first scope will also be referred to as the tag data in the first scope. The virtual tag data for virtual tags in the second scope will also be referred to as the virtual tag data in the second scope.

[0045] The first range is the range from the first position of antenna 16 to the second position of antenna 16. Here, the first position of antenna 16 is described as the position with a rotation angle of 0°. The second position of antenna 16 is described as the position with a rotation angle of 360°. In this example, the first range is the range of rotation angles from 0° to 360°. Note that the second position of antenna 16 is not limited to the position with a rotation angle of 360°. The first position of antenna 16 is an example of the first relative position of antenna 16 with respect to tag 3. The second position of antenna 16 is an example of the second relative position of antenna 16 with respect to tag 3.

[0046] The tag data in the first range is tag data for tags 3 at multiple locations of antenna 16 included in the first range. The tag data is associated with the location data of antenna 16. The multiple locations of antenna 16 may include locations at regular intervals between the first and second locations of antenna 16 included in the first range. The tag data in the first range may include tag data for all of the locations at regular intervals between the first and second locations of antenna 16. The tag data in the first range may not include tag data for some of the locations at regular intervals between the first and second locations of antenna 16.

[0047] The second range is the range from the first position of antenna 16 to the third position of antenna 16. Here, the first position of antenna 16 is assumed to be the position with a rotation angle of 0°, as described above. The third position of antenna 16 is assumed to be the position with a rotation angle of 180°. In this example, the second range is the range of rotation angles from 0° to 180°. Note that the second range is different from the first range, but if the third position of antenna 16 is the same as the second position of antenna 16, then the second range is the same as the first range. The third position of antenna 16 is not limited to the position with a rotation angle of 180°. The third position of antenna 16 is an example of the third relative position of antenna 16 with respect to tag 3.

[0048] The virtual tag data in the second range is virtual tag data relating to virtual tags at multiple locations of antenna 16 included in the second range. The virtual tag data is associated with the location data of antenna 16. The multiple locations of antenna 16 may include locations at regular intervals between the first to third locations of antenna 16 included in the second range. The tag data relating to tag 3 in the second range may include tag data for all locations at regular intervals between the first to third locations of antenna 16. The tag data relating to tag 3 in the second range may not include tag data for some of the locations at regular intervals between the first to third locations of antenna 16.

[0049] The first creation unit 212 creates virtual tag data in a second range by extracting tag data for tag 3 in a range corresponding to the width of the second range from the tag data in the first range, starting from a position shifted by a predetermined amount from the first position. The predetermined amount is a value indicating the distance between the position of tag 3 and the position of the virtual tag along the direction of movement of the antenna 16. The predetermined amount is a quantity expressed as a rotation angle. Hereinafter, the position shifted by a predetermined amount from the first position is also called the extraction start position. The range corresponding to the width of the second range is also called the extraction range. The tag data for tag 3 in the extraction range is tag data for tag 3 at multiple positions of the antenna 16 included in the extraction range. Hereinafter, the tag data for tag 3 in the extraction range is also called the tag data in the extraction range. The width of the second range is a width expressed as a rotation angle. For example, the predetermined amount is a value between 0° and 360°. The first creation unit 212 can calculate multiple predetermined amounts. The intervals between the predetermined amounts may be constant or different. For example, if the interval between predetermined quantities is constant at 90°, then the predetermined quantities are 0°, 90°, 180°, and 270°.

[0050] For example, the first creation unit 212 calculates the extraction start position. The first creation unit 212 sets the extraction start position as the starting point of the extraction range. The first creation unit 212 sets the extraction range starting from the extraction start position. The first creation unit 212 extracts tag data in the extraction range from the tag data in the first range. Extracting tag data in the extraction range from the tag data in the first range includes cutting out the tag data in the extraction range from the tag data in the first range. The first creation unit 212 shifts the range defined by the position of the antenna 16 from the extraction range to the second range for the tag data in the extraction range. The first creation unit 212 creates virtual tag data in the second range by shifting the range defined by the position of the antenna 16.

[0051] Let's explain using the example where the predetermined amount is 90°. The first position is assumed to be the position where the rotation angle is 0°. The width of the second range is assumed to be 180°. The extraction start position is the position where the rotation angle is 90°, which is 90° shifted from the position where the rotation angle is 0°. The extraction range is the range corresponding to a width of 180°, starting from the position where the rotation angle is 90°. In this case, the extraction range is the range from the position where the rotation angle is 90° to the position where the rotation angle is 270°.

[0052] The first creation unit 212 calculates the position with a rotation angle of 90° as the extraction start position. The first creation unit 212 sets the position with a rotation angle of 90° as the starting point of the extraction range. The first creation unit 212 sets the range from the position with a rotation angle of 90° to the position with a rotation angle of 270° as the extraction range. The first creation unit 212 extracts tag data from the tag data in the first range for the extraction range from the position with a rotation angle of 90° to the position with a rotation angle of 270°. The first creation unit 212 shifts the range defined by the position of the antenna 16 for the tag data in the extraction range from the extraction range from the position with a rotation angle of 90° to the position with a rotation angle of 270° to the second range from the position with a rotation angle of 0° to the position with a rotation angle of 180°. By shifting the range defined by the position of the antenna 16, the first creation unit 212 creates virtual tag data for the second range from the position with a rotation angle of 0° to the position with a rotation angle of 180°.

[0053] When the predetermined amount is 270°, the extraction range extends beyond the position where the rotation angle is 360°. Here, the tag data for tag 3 beyond the position where the rotation angle is 360° is a repetition of the tag data for tag 3 in the range from the position where the rotation angle is 0° to the position where the rotation angle is 360°. Therefore, the first creation unit 212 applies the tag data for tag 3 from the position where the rotation angle is 0° onwards to the portion of the extraction range where the rotation angle exceeds 360°.

[0054] The first creation unit 212 can create virtual tag data for multiple virtual tags based on tag data for one tag 3. For example, the first creation unit 212 can create virtual tag data for a second range for each virtual tag based on tag data for a first range for one tag 3. The positions of each of the multiple virtual tags lie on concentric circles centered on a rotation axis passing through the position of tag 3.

[0055] When the reader 1 reads multiple tags 3, the first creation unit 212 can create virtual tag data for multiple virtual tags based on the tag data for the multiple tags 3. The first creation unit 212 can create virtual tag data for multiple virtual tags for each tag 3 based on the tag data for that tag 3. For example, the first creation unit 212 can create virtual tag data for each virtual tag in a second range based on the tag data in a first range for each tag 3. The positions of the virtual tags lie on concentric circles centered on a rotation axis that passes through the positions of the tags 3 used to create the virtual tag data for these virtual tags.

[0056] For example, the training data includes virtual tag datasets for multiple virtual tags, as virtual tag data for multiple virtual tags. The virtual tag dataset includes virtual tag data in a second range created by the first creation unit 212. The virtual tag dataset is an example of virtual tag data in a second range.

[0057] The virtual tag dataset may include virtual phase data for virtual tags in a second range. The phase value changes as the position of antenna 16 changes. This is because the distance between antenna 16 and the virtual tag changes as antenna 16 moves. Since the phase value depends on the distance between antenna 16 and the virtual tag, the distribution of virtual phase data differs depending on the position of the virtual tag.

[0058] The virtual tag dataset may include virtual Doppler frequency data for virtual tags in a second range. The Doppler frequency value changes as the position of antenna 16 changes. This is because the Doppler frequency value is different when antenna 16 approaches the virtual tag and when antenna 16 moves away from the virtual tag. The distribution of virtual Doppler frequency data differs depending on the position of the virtual tag.

[0059] The virtual tag dataset may include virtual RSSI data for virtual tags in a second range. The RSSI value changes as the position of antenna 16 changes. This is because the distance between antenna 16 and the virtual tags changes as antenna 16 moves. Since the RSSI value depends on the distance between antenna 16 and the virtual tags, the distribution of virtual RSSI data differs depending on the position of the virtual tags.

[0060] The training data includes ground truth data for multiple virtual tags. The ground truth data is data indicating the location of the virtual tags. The location of the virtual tags that constitutes the ground truth data may be the coordinates indicating the location of the virtual tag, or it may be the range that contains the virtual tag. The range that contains the virtual tag is either the first judgment range or the second judgment range.

[0061] The position of the virtual tag is calculated based on the position and predetermined amount of tag 3. The first creation unit 212 calculates the position of the virtual tag based on the position and predetermined amount of tag 3. The first creation unit 212 may also calculate coordinates indicating the position of the virtual tag based on the coordinates indicating the position of tag 3 and a predetermined amount. The coordinates indicating the position of the virtual tag are coordinates shifted by a predetermined amount from the coordinates indicating the position of tag 3 on a concentric circle centered on a rotation axis passing through the coordinates indicating the position of tag 3. For example, the coordinates shifted by a predetermined amount from the coordinates indicating the position of tag 3 are coordinates shifted by a predetermined amount in the direction opposite to the direction of movement of antenna 16 from the coordinates indicating the position of tag 3. The first creation unit 212 may also calculate the range in which the virtual tag is included based on the calculated coordinates indicating the position of the virtual tag. The coordinates indicating the first determination range and the coordinates indicating the second determination range are set in advance.

[0062] Virtual phase data changes depending on the distance between antenna 16 and the virtual tag. The distribution of virtual phase data differs depending on the position of the virtual tag. There may be a certain correlation between the virtual phase data for the virtual tag at multiple positions on antenna 16 and the position of the virtual tag. Virtual Doppler frequency data differs when antenna 16 approaches the virtual tag and when antenna 16 moves away from the virtual tag. The distribution of virtual Doppler frequency data differs depending on the position of the virtual tag. There may be a certain correlation between the virtual Doppler frequency data for the virtual tag at multiple positions on antenna 16 and the position of the virtual tag. Virtual RSSI data changes depending on the distance between antenna 16 and the virtual tag. The distribution of virtual RSSI data differs depending on the position of the virtual tag. There may be a certain correlation between the virtual RSSI data for the virtual tag at multiple positions on antenna 16 and the position of the virtual tag. Thus, there may be a certain correlation between the virtual tag data for the virtual tag at multiple positions on antenna 16 and the position of the virtual tag.

[0063] The training data can be updated based on the creation of virtual tag data by the first creation unit 212.

[0064] The memory device 24 includes a learning model memory area 242. The learning model memory area 242 stores the learning model. A learning model is a model created through machine learning based on training data. Machine learning includes, but is not limited to, neural networks.

[0065] The learning model outputs judgment output data based on the input judgment data for each wireless tag to be judged read by the reader / writer 15. The judgment input data includes tag data for the wireless tag to be judged at multiple positions of the antenna 16 within the range from the starting position to a set position. The set position is a position within the movement range of the antenna 16. The set position may be a position with a rotation angle of 180°, a position with a rotation angle of 360°, or any other position. The tag data is data acquired by the reader / writer 15 in time series based on the radio waves of the wireless tag to be judged received by the reader 1. The tag data is associated with the position data of the antenna 16. The judgment output data is data concerning the range in which the wireless tag to be judged is included.

[0066] For example, data relating to the range in which the target wireless tag is included is data indicating the location of the target wireless tag. For instance, the output data for determination is data indicating that the target wireless tag is within a first determination range or data indicating that the target wireless tag is within a second determination range.

[0067] In another example, data relating to the range in which the target wireless tag is included is data indicating the level of the range in which the target wireless tag is included. The level of the range in which the target wireless tag is included is the degree to which the range indicates the probability of the target wireless tag being included. The level is explained using the example of probability, but is not limited to this. The level may be a stage selected from multiple stages, such as 10 stages. The range is either a first determination range or a second determination range. The level relating to the range in which the target wireless tag is included is explained as being a level within the first determination range, but may include levels within both the first and second determination ranges. The level within the first determination range is the level relating to the target wireless tag being within the first determination range. The level within the second determination range is the level relating to the target wireless tag being within the second determination range.

[0068] Although an example has been described in which the memory device 24 stores learning data, the explanation is not limited to this. RAM 23 may also store measurement data. In this case, RAM 23 is an example of a memory unit.

[0069] The connection interface 25 is an interface for the learning model creation device 2 to communicate with the reading device 1. The connection interface 25 may include an interface for wired connection or an interface for wireless connection.

[0070] Bus 26 includes a control bus, an address bus, and a data bus, etc. Bus 26 transmits signals exchanged between the various parts of the learning model creation device 2.

[0071] The hardware configuration of the learning model creation device 2 is not limited to the configuration described above. The learning model creation device 2 allows for the omission and modification of the above-described components, as well as the addition of new components, as appropriate.

[0072] The processor 21 implements the communication processing unit 211, the first creation unit 212, and the second creation unit 213. Each part implemented by the processor 21 can also be called a function. Each part implemented by the processor 21 can also be said to be implemented by a control unit including the processor 21, ROM 22, and RAM 23.

[0073] The communication processing unit 211 processes communication between the learning model creation device 2 and the reading device 1. The first creation unit 212 creates virtual tag data related to virtual tags. The second creation unit 213 creates a learning model using machine learning based on the training data.

[0074] The first and second judgment ranges will now be explained. Figure 2 is a top view illustrating the first determination range 81 and the second determination range 82.

[0075] The antenna 300 is positioned at the bottom of the counter stand 700. The counter stand 700 is a stand with a horizontal surface on which the tag 3 is placed.

[0076] The first determination range 81 and the second determination range 82 are horizontally separated ranges. The first determination range 81 is the range set in the central part of the surface of the counter base 700. The second determination range 82 is the range set in the outer perimeter of the surface of the counter base 700 and the range set horizontally outside the counter base 700. The second determination range 82 is set to surround the first determination range 81. In Figure 2, the second determination range 82 is set with a gap between it and the first determination range 81, without being adjacent to it. The area between the first determination range 81 and the second determination range 82 is designated as a prohibited area. The prohibited area is the area where it is forbidden to place the wireless tag to be judged.

[0077] The settings of the first determination range 81 and the second determination range 82 are not limited to these. The first determination range 81 may be a range set in the central part of the surface of the counter base 700, and the second determination range 82 may be a range set in the outer periphery of the surface of the counter base 700. The first determination range 81 may be a range set across the entire surface of the counter base 700, and the second determination range 82 may be a range set horizontally outside the counter base 700. The second determination range 82 is not limited to a range set to surround the first determination range 81.

[0078] This section describes how to read tag 3 using reader device 1.

[0079] Figures 3 and 4 are top views illustrating the movement of antenna 16. Figure 5 is a side view illustrating the movement of antenna 16. The antenna 16 moves along the direction of movement 4 in the horizontal plane. The direction of movement 4 is the circumferential direction centered on the rotation axis 5 in the vertical direction. The concentric circles 6 are concentric circles centered on the rotation axis 5 and passing through the positions of the tags 3 placed on the counter base 700. The reading device 1 may read the tag 3 and acquire tag data each time the tag 3 is rotated and its orientation is changed, as shown in Figure 4, even if the tag 3 is in the same position. The reading processing unit 111 may read multiple tags 3 that are in the same position in the horizontal plane but are located at different heights, and acquire tag data for each tag 3.

[0080] This section explains an example of creating virtual tag data based on existing tag data. Figure 6 is a diagram illustrating the creation of virtual tag data based on tag data. Figure 6 shows a top view of the counter 700.

[0081] The concentric circles 7 are centered on the axis of rotation 5 and pass through positions PA and PB on the horizontal surface of the counter base 700. The distance between positions PA and PB along the direction of movement 4 of the antenna 16 is 90°.

[0082] Figure 7 is a graph showing tag data for location PA tag A. Here, tag 3, placed at position PA as shown in Figure 6, is referred to as tag A. The horizontal axis represents the position of antenna 16 as the rotation angle of antenna 16. The vertical axis represents the phase. The reading device 1 is assumed to have acquired tag data related to tag A in accordance with the movement of the antenna 16 based on the reading of tag A.

[0083] Figure 8 is a graph showing tag data for tag B at location PB. Here, tag 3, placed at position PB as shown in Figure 6, will be referred to as tag B. The horizontal axis represents the position of antenna 16 as the rotation angle of antenna 16. The vertical axis represents the phase. The reading device 1 is assumed to have acquired tag data related to tag B in accordance with the movement of antenna 16 based on the reading of tag B.

[0084] Comparing Figure 7 and Figure 8, the graphs of tag data for tag A and tag data for tag B are shifted by 90°, similar to the interval between location PA and location PB. Therefore, the graph of tag data for tag B is similar to the graph of tag data for tag A shifted by 90°.

[0085] Figure 9 is a graph showing virtual tag data for virtual tag b of location PB. Here, the virtual tag assumed to be located at position PB as shown in Figure 6 is referred to as virtual tag b. The horizontal axis represents the position of antenna 16 as the rotation angle of antenna 16. The vertical axis represents the phase.

[0086] The learning model creation device 2 is assumed to have created virtual tag data for virtual tag b in the second range based on the tag data for tag A in the first range. The first range is assumed to be a range of rotation angles from 0° to 360°, as shown in Figure 7. The second range is assumed to be a range of rotation angles from 0° to 360°, as shown in Figure 9. The predetermined amount is 90°, which represents the distance between position PA and position PB. The extraction start position is the position with a rotation angle of 90°, which is a 90° shift from the position with a rotation angle of 0°.

[0087] The graph of virtual tag data for virtual tag b shown in Figure 9 is a graph of tag data for tag A shown in Figure 7, shifted by 90°. Specifically, the virtual tag data for virtual tag b in the rotation angle range of 0° to 270° corresponds to the tag data for tag A in the rotation angle range of 90° to 360°. The virtual tag data for virtual tag b in the rotation angle range of 270° to 360° corresponds to the tag data for tag A in the rotation angle range of 0° to 90°.

[0088] Comparing Figure 8 and Figure 9, it can be seen that the graph of tag data for tag B and the graph of virtual tag data for virtual tag b are similar. Thus, the learning model creation device 2 can appropriately create virtual tag data for virtual tag b, which is assumed to be at location PB, based on the tag data for tag A.

[0089] Figure 10 shows the relationship between tag data for location PA tag A and virtual tag data for location PB virtual tag b. Figure 10 shows the tag data for tag A and virtual tag b for each position of antenna 16, indicated by the rotation angle of antenna 16. 'a' represents the interval for acquiring tag data, indicated by the rotation angle of antenna 16. Thus, the arrangement of virtual tag data for tag b is a shifted version of the arrangement of tag data for tag A.

[0090] (Example of operation) This section describes the processing in the learning model creation system S. The processing procedure described below is merely an example, and each process may be modified as much as possible. Furthermore, depending on the embodiment, steps in the processing procedure described below may be omitted, replaced, or added as appropriate.

[0091] The processing in the reading device 1 will be explained below. Figure 11 is a flowchart showing an example of processing by the processor 11 of the reading device 1. For example, one or more tags 3 are placed on the counter 700. The processor 11 of the reader 1 may start processing based on a start instruction entered by the user.

[0092] The reading processing unit 111 determines whether or not the antenna 16 is present at the starting position (ACT1). If the antenna 16 is not present at the starting position (ACT1, NO), the process transitions from ACT1 to ACT2. If the antenna 16 is present at the starting position (ACT1, YES), the process transitions from ACT1 to ACT3.

[0093] The reading processing unit 111 controls the antenna 16 to move to the starting position (ACT2). In ACT2, for example, the reading processing unit 111 controls the drive unit 17 to move the antenna 16 to the starting position.

[0094] The reading processing unit 111 controls the movement of the antenna 16 (ACT3). In ACT3, for example, the reading processing unit 111 controls the drive unit 17 to move the antenna 16 from the starting position to the ending position.

[0095] The reading processing unit 111 controls the start of radio wave transmission from the antenna 16 (ACT4). In ACT4, for example, the reading processing unit 111 controls the start of radio wave transmission from the antenna 16 based on the start of movement of the antenna 16 from its starting position. The antenna 16 starts transmitting radio waves to read the identification information of the tag 3 from the tag 3.

[0096] The reading processing unit 111 acquires the identification information and tag data obtained by the reader / writer 15 for each tag 3 read by the reader / writer 15 (ACT5). If the reading processing unit 111 acquires the identification information and tag data (ACT5, YES), the process transitions from ACT5 to ACT6. If the reading processing unit 111 does not acquire the identification information and tag data (ACT5, NO), the process transitions from ACT5 to ACT7.

[0097] Based on the acquisition of identification information and tag data, the reading processing unit 111 stores the tag data relating to tag 3 identified by the identification information in the measurement data storage area 181 (ACT6).

[0098] The reading processing unit 111 determines whether the movement of the antenna 16 has finished (ACT7). In ACT7, for example, the reading processing unit 111 determines whether the movement of the antenna 16 from the start position to the end position has finished. If the movement of the antenna 16 has finished (ACT7, YES), the process transitions from ACT7 to ACT8. If the movement of the antenna 16 has not finished (ACT7, NO), the process transitions from ACT7 to ACT5.

[0099] The reading processing unit 111 repeats the processes of ACT5 and ACT6 from the time the antenna 16 starts moving at the starting position until it finishes moving at the ending position.

[0100] In ACT5, the reading processing unit 111 acquires identification information and multiple tag data at multiple locations on the antenna 16 for each tag 3 read by the reader / writer 15. For example, the reading processing unit 111 can acquire identification information and multiple phase data, Doppler frequency data, or RSSI data at multiple locations on the antenna 16 for each tag 3.

[0101] In ACT6, the reading processing unit 111 stores multiple tag data at multiple locations on the antenna 16 in the measurement data storage area 181 for each tag 3 read by the reader / writer 15. For example, the reading processing unit 111 can store multiple phase data, Doppler frequency data, or RSSI data at multiple locations on the antenna 16 in the measurement data storage area 181 for each tag 3.

[0102] The reading processing unit 111 controls the termination of radio wave transmission from the antenna 16 (ACT8). In ACT8, for example, the reading processing unit 111 controls the termination of radio wave transmission from the antenna 16 based on the termination of the movement of the antenna 16. The termination of the movement of the antenna 16 is the termination of the movement of the antenna 16 from the starting position to the ending position. The antenna 16 terminates the radio wave transmission for reading the identification information of the tag 3.

[0103] The communication processing unit 112 transmits the measurement data to the learning model creation device 2 (ACT9). In ACT9, for example, the communication processing unit 112 transmits the measurement data to the learning model creation device 2 via the connection interface 14.

[0104] The processing in the learning model creation device 2 will be explained. The meanings of the terms used in the following explanation are as follows: The division angle θ0 is a value that indicates how often tag data for a virtual tag is created. The division angle θ0 is used to calculate the predetermined quantity described above. The predetermined quantity is a value that is 0 times or a positive integer multiple of the division angle θ0. For example, if the division angle θ0 is 90°, the predetermined quantities are 0°, 90°, 180°, and 270°. The data extraction start angle θ1 is a value that indicates the rotation angle at which the extraction of tag data from the tag data in the first range to the tag data in the extraction range begins. The data extraction start angle θ1 is an example of the extraction start position described above. The data extraction angle θ2 is a value that indicates the range of degrees over which virtual tag data will be created. The data extraction angle θ2 is an example of the width of the second range. The range θ1 to θ1+θ2 is an example of an extraction range. The number of data points n is a value that indicates the number of data points that are valid for use in machine learning. If the number of tag data points in the extracted range is n or greater, the number of virtual tag data points in the second range based on the tag data points in the extracted range is appropriate. On the other hand, if the number of tag data points in the extracted range is less than n, the number of virtual tag data points in the second range based on the tag data points in the extracted range is not appropriate.

[0105] Figure 12 is a flowchart showing an example of processing by the processor 21 of the learning model creation device 2.

[0106] The communication processing unit 211 acquires measurement data from the reading device 1 (ACT11). In ACT11, for example, the communication processing unit 211 acquires measurement data from the reading device 1 via the connection interface 25.

[0107] The first creation unit 212 sets the division angle θ0 (ACT12). The division angle θ0 may be stored in the storage device 24. The division angle θ0 can be set as appropriate.

[0108] The first creation unit 212 sets the data extraction angle θ2 (ACT13). The data extraction angle θ2 may be stored in the storage device 24. The data extraction angle θ2 can be set as appropriate.

[0109] The first creation unit 212 sets the number of data points n (ACT14). The number of data points n may be stored in the storage device 24. The number of data points n can be set as appropriate.

[0110] The first creation unit 212 processes data synthesis (ACT15). In ACT15, for example, the first creation unit 212 creates tag data for a first range for each tag 3, based on the tag data for the movement range of the antenna 16 included in the measurement data. An example of the data synthesis process will be described later.

[0111] The first creation unit 212 processes data shifts (ACT16). In ACT16, for example, the first creation unit 212 creates virtual tag data for each tag 3 based on the tag data for tag 3. An example of the data shift process will be described later.

[0112] The first creation unit 212 determines whether it has completed the creation of virtual tag data for virtual tags based on the tag data for all tags 3 (ACT17). If creation is complete (ACT17, YES), the process transitions from ACT17 to ACT18. If creation is not complete (ACT17, NO), the process transitions from ACT17 to ACT16.

[0113] The second creation unit 213 creates a learning model (ACT18) using machine learning based on training data that includes virtual tag data for multiple virtual tags. The second creation unit 213 stores the created learning model in the learning model memory area 242.

[0114] Figure 13 is a flowchart showing an example of the data synthesis process performed by the processor 21 of the learning model creation device 2. The processor 21 processes the data synthesis shown in Figure 13 for each tag 3.

[0115] The first creation unit 212 synthesizes the tag data within the movement range of the antenna 16 (ACT 151). When the reader 1 reads the tag 3 using multiple antennas 16, it acquires tag data within the movement range of each antenna 16. In ACT 151, for example, the first creation unit 212 synthesizes the tag data acquired for each antenna 16 within the movement range of the antenna 16. Based on the synthesis of the tag data within the movement range of the antenna 16, the first creation unit 212 creates tag data within a first range.

[0116] The first creation unit 212 determines whether it has finished synthesizing the tag data for all antennas 16 within the range of movement of the antennas 16 (ACT152). If the synthesis is complete (ACT152, YES), the process ends. If the synthesis is not complete (ACT152, NO), the process transitions from ACT152 to ACT151.

[0117] Figure 14 is a flowchart showing an example of data shift processing by the processor 21 of the learning model creation device 2. The processor 21 processes the data shift shown in Figure 14 for each tag 3.

[0118] The first creation unit 212 calculates the data extraction start angle θ1 (ACT161). In ACT161, for example, the first creation unit 212 calculates a predetermined amount based on the division angle θ0. Based on the calculated predetermined amount, the first creation unit 212 calculates the data extraction start angle θ1, which is shifted by a predetermined amount from the first position.

[0119] The first creation unit 212 calculates coordinates indicating the position of the virtual tag based on the position and predetermined quantity of the tag 3 included in the measurement data (ACT162).

[0120] The first creation unit 212 determines whether the location of the virtual tag is within a prohibited area (ACT163). In ACT163, for example, the first creation unit 212 compares the coordinates indicating the location of the virtual tag with the coordinates indicating the prohibited area. The coordinates indicating the prohibited area are assumed to be pre-set.

[0121] If the virtual tag is not located in a prohibited area (ACT163, YES), the process transitions from ACT163 to ACT164. In this case, the first creation unit 212 creates virtual tag data for this virtual tag. If the virtual tag is located in a prohibited area (ACT163, NO), the process transitions from ACT163 to ACT168. In this case, the first creation unit 212 does not create virtual tag data for this virtual tag.

[0122] The first creation unit 212 extracts tag data in the range θ1 to θ1+θ2 (ACT164). In ACT164, for example, the first creation unit 212 sets the extraction range as θ1 to θ1+θ2 based on the data extraction start angle θ1 and the data extraction angle θ2. The first creation unit 212 extracts tag data in the range θ1 to θ1+θ2 set as the extraction range from the tag data in the first range.

[0123] The first generation unit 212 counts the number of tag data in the range θ1 to θ1+θ2 (ACT165).

[0124] The first creation unit 212 compares the counted number of data points with the number of data points n (ACT166). If the counted number of data points is greater than or equal to the number of data points n (ACT166, YES), the process transitions from ACT166 to ACT167. In this case, the first creation unit 212 creates virtual tag data related to the virtual tag. If the counted number of data points is less than the number of data points n (ACT166, NO), the process transitions from ACT166 to ACT168. In this case, the first creation unit 212 does not create virtual tag data related to the virtual tag.

[0125] The first creation unit 212 stores data for the virtual tag in the training data storage area 241 (ACT167). In ACT167, for example, the first creation unit 212 creates virtual tag data in a second range based on the tag data in the range θ1 to θ1+θ2. The first creation unit 212 stores the created virtual tag data in the second range in the training data storage area 241. Based on the coordinates calculated in ACT162, the first creation unit 212 stores data indicating the position of the virtual tag as ground truth data in the training data storage area 241.

[0126] The first creation unit 212 determines whether the data extraction start angle θ1 is greater than 360° (ACT168). If the data extraction start angle θ1 is greater than 360° (ACT168, YES), the process ends. If the data extraction start angle θ1 is 360° or less (ACT168, NO), the process transitions from ACT168 to ACT161.

[0127] Figure 15 is a diagram illustrating the creation of virtual tag data by the processor 21 of the learning model creation device 2. Figure 15 shows tag data for a tag 3 placed at an arbitrary position, within a rotation angle range from 0° to 360°. The horizontal axis represents the position of the antenna 16 as its rotation angle. The vertical axis represents the phase. Assume that the reader 1 is able to acquire tag data within the rotation angle range of 0° to 180°, but not within the rotation angle range of 180° to 360°. Assume that the division angle θ0 is 90°. Assume that the data extraction angle θ2 is 180°.

[0128] Let's describe the virtual tag at the same position as tag 3. In this case, the data extraction start angle θ1 is 0°. The extraction range is the range from 0° to 180° shown by the solid line. The number of tag data in the range from 0° to 180° is n or more. The learning model creation device 2 creates virtual tag data in the range from 0° to 180° based on the tag data in the range from 0° to 180° for this virtual tag.

[0129] This section describes a virtual tag located 90° shifted from the position of tag 3. In this case, the data extraction start angle θ1 is 90°. The extraction range is the 90° to 170° range shown by the dashed line. The number of tag data in the 90° to 270° range is n or more. The learning model creation device 2 creates virtual tag data in the 0° to 180° range for this virtual tag, based on the tag data in the 90° to 270° range.

[0130] This section describes a virtual tag located 180° shifted from the position of tag 3. In this case, the data extraction start angle θ1 is 180°. The extraction range is the 180° to 360° range indicated by the dashed line. The number of tag data points in the 180° to 360° range is less than the data point number n. The learning model creation device 2 does not create virtual tag data for this virtual tag.

[0131] (modified version) Let me explain the first variation. The reading device 1 is not limited to including one antenna 16. The reading device 1 may include multiple antennas 16.

[0132] Figure 16 is a top view illustrating a modified example of the reading device 1. Here, the reading device 1 includes two antennas 16. One antenna 16 moves in the horizontal plane along a movement direction 41. The movement direction 41 is the circumferential direction around the rotation axis 5. The other antenna 16 moves in the horizontal plane along a movement direction 42. The movement direction 42 is the circumferential direction around the rotation axis 5. Each of the two antennas 16 rotates 180° along the circumferential direction around the rotation axis 5, but the range of movement of each of the two antennas 16 is different.

[0133] Let me explain the second variation. The reading device 1 is not limited to moving the antenna 16 in a circular direction around a rotation axis aligned with the vertical. The direction of movement of the antenna 16 may also be linear. Figure 17 is a side view illustrating a second modified example of the reading device 1. Figure 18 is a top view illustrating a second modified example of the reading device 1. The antenna 16 moves along the direction of movement 43. The direction of movement 43 is a straight horizontal direction. In the second modification, the first creation unit 212 sets the position of the virtual tag 30 to a position symmetric to the position of tag 3 with respect to the trajectory of the antenna 16 moving along the direction of movement 43. The first creation unit 212 creates virtual tag data for the virtual tag 30 based on the tag data for tag 3. For example, the first creation unit 212 creates virtual tag data for the virtual tag 30 in a second range based on the tag data for tag 3 in a first range. Since the position of the virtual tag 30 is symmetric to the position of tag 3, the virtual tag data for the virtual tag 30 in the second range is the same as the tag data for tag 3 in the first range. The first range and the second range are the same range defined by the position of the antenna 16.

[0134] Let me explain the third variation. The reader 1 moves the antenna 16, but is not limited to this. The position of the antenna 16 is fixed, and the reader 1 may move the tag 3. In this example, the reader 1 may move the stage on which the tag 3 is placed. Moving the tag 3 is one example of changing the relative position of the antenna 16 with respect to the tag 3. The reader 1 may also move both the antenna 16 and the tag 3. Moving both the antenna 16 and the tag 3 is one example of changing the relative position of the antenna 16 with respect to the tag 3.

[0135] (principle / phenomenon) As described above, the ability of the learning model creation device 2 to create virtual tag data for virtual tags based on tag data for tag 3 will be explained in terms of the relationship between the position of tag 3 and the tag data for tag 3.

[0136] Figure 19 is a top view illustrating the arrangement of multiple tags. Antenna 16 moves in the horizontal plane along the direction of movement 4. The direction of movement 4 is the circumferential direction centered on the rotation axis 5 along the vertical direction. The concentric circles 6 are centered on the rotation axis 5 and pass through the positions of tag (1), tag (2), tag (3), and tag (4) placed on the counter base 700. The positions of tag (1), tag (2), tag (3), and tag (4) are 90° apart from each other with respect to the rotation axis 5. Tags (1), (2), (3), and (4) are wireless tags configured in the same way as tag 3 described above.

[0137] Figure 20 is a graph showing tag data. The graph shows tag data acquired by rotating antenna 16 360° around rotation axis 5. The horizontal axis represents the position of antenna 16 as the rotation angle of antenna 16. The vertical axis represents the phase. Figure 21 shows graphs of tag data for tag (1), tag data for tag (2), tag data for tag (3), and tag data for tag (4). Each graph shows tag data acquired by rotating antenna 16 180° around rotation axis 5. The horizontal axis of each graph indicates the position of antenna 16 as the rotation angle of antenna 16. The vertical axis of each graph indicates the phase.

[0138] Comparing Figure 20 and Figure 21, it can be seen that each graph shown in Figure 21 can be created by shifting the graph shown in Figure 20 along the horizontal axis and cropping the necessary range (in this case, a range with a width of 180°) from the graph shown in Figure 20.

[0139] Figure 22 is a graph showing tag data for tag (1) and tag data for tag (4). The graph shows tag data acquired by rotating antenna 16 360° around rotation axis 5. The horizontal axis represents the position of antenna 16 as the rotation angle of antenna 16. The vertical axis represents the phase. The graphs of tag data for tag (1) and tag data for tag (4) are shifted by 90°, similar to the distance between the positions of tag (1) and tag (4).

[0140] (effect) According to one embodiment, the learning model creation device 2 can create virtual tag data for virtual tags based on tag data for tag 3. As a result, the learning model creation device 2 can create virtual tag data for multiple virtual tags based on the tag data for one tag 3. The number of coordinates for placing tag 3, which is read by the reader 1, is reduced. Therefore, the learning model creation device 2 can easily acquire the training data used to create the learning model. Furthermore, the learning model creation device 2 can reduce the time required to acquire the training data used to create a learning model with high judgment accuracy.

[0141] (Other embodiments) The reading device may be implemented as a single device, as explained in the example above, or as multiple devices with distributed functions. The learning model creation device may be implemented as a single device, as explained in the example above, or as multiple devices with distributed functions.

[0142] The embodiments described above may apply not only to the apparatus but also to the methods performed by the apparatus. The embodiments described above may apply to a program that can cause the computer of the apparatus to perform each function. The embodiments described above may apply to a recording medium that stores the program. The embodiments described above may apply not only to the system but also to the methods performed by multiple elements included in the system.

[0143] A processing circuit includes one or more circuits that perform multiple processes through multiple functions. For example, the circuit may be, but is not limited to, a processor, an ASIC (Application Specific Integrated Circuit), or an FPGA (Field-Programmable Gate Array).

[0144] Each of the one or more circuits that make up a processing circuit performs one or more of the multiple processes. If the processing circuit consists of a single circuit, the single circuit performs all of the multiple processes. If the processing circuit consists of multiple circuits, each of the multiple circuits performs some of the multiple processes. Some of the multiple processes may be one of the multiple processes, or two or more of the multiple processes. If the processing circuit consists of multiple circuits, the multiple circuits may be contained in a single device, or they may be distributed across multiple devices.

[0145] The program may be transferred while stored in the device, or it may be transferred without being stored in the device. In the latter case, the program may be transferred via a network, or it may be transferred while recorded on a recording medium. The recording medium is a non-temporary tangible medium. The recording medium is a computer-readable medium. The recording medium can be any medium that is capable of storing a program and is readable by a computer, such as a CD-ROM or memory card, and its form is not limited.

[0146] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents.

[0147] The above-described embodiment may also be expressed as follows: (1) A first creation unit that creates virtual tag data for a virtual tag based on tag data for a wireless tag, A second creation unit creates a learning model for determining the location of a target wireless tag based on tag data relating to a target wireless tag, using machine learning based on learning data including virtual tag data relating to multiple virtual wireless tags created by the first creation unit. A learning model creation device equipped with the following features. (2) The learning model creation apparatus according to (1), wherein the first creation unit creates virtual tag data for the virtual tag in a second range from the first relative position to the third relative position of the antenna, based on tag data for the wireless tag in a first range from the first relative position to the antenna communicating with the wireless tag to the second relative position of the antenna. (3) The learning model creation device according to (2), wherein the first creation unit creates virtual tag data for the virtual tag in the second range by extracting tag data for the wireless tag in a range corresponding to the width of the second range from tag data for the wireless tag in the first range, starting from a relative position shifted by a predetermined amount from the first relative position. (4) The learning model creation device according to (3), wherein the first creation unit calculates the position of the virtual wireless tag based on the position of the wireless tag and the predetermined amount. (5) A reading device and A learning model creation device, Equipped with, The aforementioned reading device, An antenna that communicates with wireless tags, A drive unit that moves the relative position of the antenna with respect to the wireless tag, An acquisition unit that acquires tag data relating to the wireless tag at multiple relative positions of the antenna, Equipped with, The aforementioned learning model creation device is A first creation unit that creates virtual tag data for a virtual tag based on the tag data for the aforementioned wireless tag, A second creation unit creates a learning model for determining the location of a target wireless tag based on tag data relating to a target wireless tag, using machine learning based on learning data including virtual tag data relating to multiple virtual wireless tags created by the first creation unit. Equipped with, A system for creating learning models. (6) To the computer, A function to create virtual tag data for virtual tags based on tag data for wireless tags, A function to create a learning model for determining the location of a target wireless tag based on tag data related to a target wireless tag, using machine learning based on training data that includes virtual tag data for multiple virtual wireless tags created. A program to make it executable. [Explanation of symbols]

[0148] 1...Reader, 2...Learning model creation device, 3...Tag, 4...Direction of movement, 5...Rotation axis, 6...Concentric circles, 7...Concentric circles, 11...Processor, 12...ROM, 13...RAM, 14...Connection interface, 15...Reader / writer, 16...Antenna, 17...Drive unit, 18...Storage device, 19...Bus, 21...Processor, 22...ROM, 23...RAM, 24...Storage device, 25...Connection interface, 26...Bus, 3 0...Virtual tag, 41...Direction of movement, 42...Direction of movement, 43...Direction of movement, 81...First judgment range, 82...Second judgment range, 111...Reading processing unit, 112...Communication processing unit, 181...Measurement data storage area, 211...Communication processing unit, 212...First creation unit, 213...Second creation unit, 241...Learning data storage area, 242...Learning model storage area, 700...Counter stand, PA...Position, PB...Position, S...Learning model creation system.

Claims

1. A first creation unit that creates virtual tag data for a virtual tag in a second range by starting from a relative position shifted by a predetermined amount from a first relative position of an antenna communicating with a wireless tag, and extracting tag data for the wireless tag in a range corresponding to the width of a second range from the first relative position to a third relative position of the antenna from the tag data for the wireless tag in a first range from the first relative position to a second relative position of the antenna, A second creation unit creates a learning model for determining the location of a target wireless tag based on tag data relating to a target wireless tag, using machine learning based on learning data including virtual tag data relating to a plurality of virtual wireless tags created by the first creation unit. A learning model creation device equipped with the following features.

2. The learning model creation device according to claim 1, wherein the first creation unit calculates the position of the virtual wireless tag based on the position of the wireless tag and the predetermined amount.

3. A reader device for acquiring tag data relating to the wireless tag, A learning model creation device according to claim 1 or claim 2, A learning model creation system equipped with the following features.

4. On the computer, A function to create virtual tag data for a virtual tag in a second range by starting from a relative position shifted by a predetermined amount from a first relative position of the antenna communicating with the wireless tag, and extracting tag data for the wireless tag in a range corresponding to the width of a second range from the first relative position to the third relative position of the antenna, from the tag data for the wireless tag in a first range from the first relative position to the second relative position of the antenna, and A function to create a learning model for determining the location of a target wireless tag based on tag data related to a target wireless tag, using machine learning based on training data that includes virtual tag data for multiple virtual wireless tags created. A program to make it executable.

5. The computer, The program according to claim 4, which further enables the function of calculating the position of the virtual radio tag based on the position of the radio tag and the predetermined amount.

Images