JP7688286B2 - Position estimation device, position estimation method, and position estimation program - Google Patents

Description

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

The Internet of Things (IOT), in which various devices are connected to the Internet, is becoming a reality. Various devices, such as automobiles, drones, and construction machinery vehicles, are becoming wirelessly connected. Regarding wireless communication standards, the wireless LAN (Local Area Network) defined by the standard IEEE 802.11, Bluetooth (registered trademark), cellular communication using LTE and 5G, LPWA (Low Power Wide Area) communication for IOT, ETC (Electronic Toll Collection System), VICS (Vehicle Information and Communication System), ARIB-STD-T109, and other wireless communication standards that support them are also developing and are expected to become more widespread in the future.

In order to ensure high throughput and reliability, wireless communication devices have adopted MIMO (Multiple input multiple output) communication technology using multiple antennas. MIMO communication technology can improve throughput and reliability by utilizing channel information that indicates how radio waves are propagating between the transmitting and receiving sides. For example, wireless communication devices on the transmitting side support a function for transmitting a feedback signal that conveys channel information to wireless communication devices on the receiving side (see Non-Patent Document 1).

There is also known technology that uses channel information related to radio wave propagation to estimate the position of a wireless communication device (see Non-Patent Documents 2 and 3). For example, the position of a wireless communication device is identified based on the arrival time and level of wireless signals wirelessly communicated with multiple base stations.

“Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications”, IEEE Computer Society, IEEE Std 802.11, 2016, p.2396-p.2400 Y. Tao and others, “A Novel System for WiFi Radio Map Automatic Adaptation and Indoor Positioning”, in IEEE Transactions on Vehicular Technology, vol. 67, no. 11, November 2018, p.10683-p.10692 H. CAO, 5 others, “Indoor Positioning Method Using WiFi RTT Based on LOS Identification and Range Calibration”, in Proc., ISPRS International Journal of Geo-Information, September 2020

However, conventional position estimation technologies had the problem of requiring wireless communication devices to communicate wirelessly with multiple base stations simultaneously and with high time resolution capabilities, resulting in high costs for estimating the position of an object.

The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a technology that can estimate the position of an object at low cost.

A position estimation device of one embodiment of the present invention comprises a wireless communication unit that receives a wireless signal transmitted from a wireless communication unit of a fixed terminal and acquires channel information related to radio wave propagation from the wireless signal, an input feature generation unit that converts the channel information into input features that can be input to a position estimation model, and a position estimation model utilization unit that inputs the input features into a position estimation model that models, through machine learning, the relationship between the channel information related to radio wave propagation and position information of a position estimation target that is located in the same environment as the fixed terminal, thereby estimating and calculating the position of the position estimation target.

A position estimation method according to one aspect of the present invention is a position estimation method performed by a position estimation device, which includes the steps of receiving a wireless signal transmitted from a wireless communication unit of a fixed terminal and acquiring channel information related to radio wave propagation from the wireless signal, converting the channel information into input features that can be input to a position estimation model, and inputting the input features into a position estimation model that models, by machine learning, the relationship between the channel information related to radio wave propagation and position information of a position estimation target located in the same environment as the fixed terminal, thereby estimating and calculating the position of the position estimation target.

A position estimation program according to one aspect of the present invention causes a computer to function as the position estimation device described above.

The present invention provides technology that can estimate the position of an object at low cost.

FIG. 1 is a diagram showing the overall configuration (first example) of a wireless communication system according to this embodiment. FIG. 2 is a diagram showing an overall configuration (second example) of a wireless communication system according to this embodiment. FIG. 3 is a diagram showing a configuration (variation) of the position estimation model utilization unit. FIG. 4 is a diagram showing the operation (first example) of the wireless communication system. FIG. 5 is a diagram showing the operation (second example) of the wireless communication system. FIG. 6 is a diagram showing the experimental environment area. FIG. 7 is a diagram showing an example of division of the experiment environment area. FIG. 8 shows the measurement results of the position estimation error (when only the radio signals of the fixed terminals are used). FIG. 9 is a diagram showing the measurement result of the position estimation error (when the radio signal of the position measurement target is further used). FIG. 10 is a diagram showing measurement results of position estimation. FIG. 11 is a diagram illustrating a hardware configuration of the position estimation device.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same parts are given the same reference numerals and the description will be omitted.

Summary of the Invention
The present invention estimates the position of a specific object by using commonly used general-purpose wireless communication information, that is, by using channel information related to radio wave propagation included in a wireless signal of a wireless communication terminal and a position estimation model. Specifically, a position estimation model is generated in advance by modeling the relationship between the channel information included in the wireless signal and the position information of a specific object by machine learning, and channel information acquired from the wireless signal (input feature converted so that the channel information can be input to the position estimation model) is input to the position estimation model, thereby estimating the position of the specific object in the real world at the time of position estimation.

In this way, the present invention estimates the position of a specific object using channel information regarding radio wave propagation contained in the wireless signal of a wireless communication terminal and a position estimation model, making it possible to estimate the position of a specific object using general-purpose wireless communication, and providing a technology that can estimate the position of a specific object at low cost.

Channel information is information about how radio waves propagate between a transmitting wireless communication terminal and a receiving wireless communication terminal, and about the communication quality of wireless communication. For example, it is information about the received power and radio wave propagation coefficient in wireless communication, the channel matrix that represents the state of radio wave propagation between multiple antennas equipped on a transmitting wireless communication terminal and multiple antennas equipped on a receiving wireless communication terminal in MIMO (Multiple Input Multiple Output) communication technology, and the signal-to-noise-interference power.

The input features are the features of channel information that have been converted so that they can be input to the position estimation model. For example, the input features are the unconverted channel information itself, or numerical values obtained by performing various calculations on the channel information.

A specific object is a movable target for position estimation that is located in the same environment as the wireless communication terminal. The position of a specific object is, for example, the position on the path along which the specific object is moving, the position in two-dimensional space (such as a map), or the position in three-dimensional space. In addition to this position information, more detailed physical conditions such as orientation and speed may also be estimated.

[Overall configuration of a wireless communication system (first example)]
FIG. 1 is a diagram showing the overall configuration (first example) of a wireless communication system according to this embodiment.

The first example wireless communication system comprises a position estimation device 1, fixed terminals 3-1 to 3-M, and position estimation targets 2-1 to 2-P.

The position estimation device 1 collects channel information of a wireless signal transmitted from at least one of the fixed terminals 3-1 to 3-M, and estimates the position of a specific object located in the same environment as the fixed terminals 3-1 to 3-M, i.e., the position of at least one of the multiple position estimation targets 2-1 to 2-P.

The fixed terminals 3-1 to 3-M are equipped with wireless communication units 3-1-1 to 3-M-1. The wireless communication units 3-1-1 to 3-M-1 transmit wireless signals including pilot signals that are known during transmission and reception, or channel information between any wireless communication unit. The any wireless communication unit is the wireless communication units 1-1 to 1-R provided in the position estimation device 1, or any other wireless communication unit.

The position estimation device 1 receives the wireless signals via the wireless communication units 1-1 to 1-R. From the received wireless signals, the position estimation device 1 acquires channel information between any wireless communication unit 3-i-1 (1≦i≦M) of the multiple fixed terminals 3-1 to 3-M and any of its own wireless communication units 1-1 to 1-R. The position estimation device 1 then inputs the acquired channel information to the input feature generation unit 1-2. The input feature generation unit 1-2 converts the channel information into input features suitable for input to the position estimation model, and inputs the converted input features to the position estimation model utilization unit 1-3.

Thereafter, the location estimation model utilization unit 1-3 inputs the input features of channel information collected from at least one fixed terminal 3-i (1≦i≦M) among the multiple fixed terminals 3-1 to 3-M into a location estimation model that models the relationship between the location information of the location estimation target and the input features of the channel information using machine learning, thereby estimating the location of a moving or stopped location estimation target 2-j (1≦j≦P).

[Fixed terminal configuration]
The fixed terminal 3-i is a wireless communication terminal equipped with a wireless communication unit 3-i-1 and installed in a predetermined environment. The fixed terminal 3-i is preferably fixed to a wall, floor, ceiling, etc., so as not to move. The fixed terminal 3-i may be realized using a special dedicated device, or may be realized using any terminal having a built-in wireless communication unit, such as a smartphone, PC, or dongle. One fixed terminal 3-i may be equipped with multiple wireless communication units 3-i-1. There may be multiple fixed terminals 3-i.

[Configuration of the position estimation device]
The position estimation device 1 is, for example, a base station installed at a major location. The position estimation device 1 may have any configuration including a wireless communication unit capable of communicating with the fixed terminal 3-i, or a wireless communication unit capable of decoding a wireless signal transmitted from the fixed terminal 3-i. The position estimation device 1 does not necessarily need to have a function of transmitting a wireless signal, and may be a device that only receives the signal.

As shown in FIG. 1, the location estimation device 1 includes, for example, wireless communication units 1-1 to 1-R, an input feature generation unit 1-2, a location estimation model utilization unit 1-3, and a location estimation model training unit 1-4.

The wireless communication units 1-1 to 1-R are communication units that perform wireless communication or receive wireless signals. The wireless communication units 1-1 to 1-R may be compatible with multiple frequencies, multiple frequency bands, or multiple wireless communication systems. In FIG. 1, the wireless communication units 1-1 to 1-R are shown configured to receive wireless signals transmitted from M fixed terminals 3-1 to 3-M.

Any of the wireless communication units 1-1 to 1-R may be a base station that communicates with the wireless communication units 3-1-1 to 3-M-1 of the fixed terminals 3-1 to 3-M. The fixed terminals 3-1 to 3-M may communicate with any base station not shown in FIG. 1 and receive wireless signals that they themselves transmit to that base station. A mixed configuration of these is also possible. Since the wireless communication units 1-1 to 1-R on the receiving side are capable of decoding wireless signals at a single carrier frequency, the fixed terminals 3-1 to 3-M may reduce the complexity of the circuits of the wireless communication units 3-1-1 to 3-M-1 by transmitting wireless signals at the same frequency.

The wireless communication units 1-1 to 1-R each have a function of receiving a wireless signal and acquiring, from the received wireless signal, channel information regarding radio wave propagation between the fixed terminal 3-i and the position estimation device 1, or channel information regarding radio wave propagation between a wireless communication device other than the fixed terminal 3-i and the position estimation device 1. When the wireless communication units 1-1 to 1-R acquire channel information corresponding to the fixed terminal 3-i, they input the acquired channel information to the input feature generation unit 1-2.

The input feature generator 1-2 has a function of converting the input channel information into input features that can be input to the position estimation model. The input feature generator 1-2 inputs the converted input features of the channel information to the position estimation model utilization unit 1-3.

The position estimation model utilization unit 1-3 has a function of estimating and calculating the position of at least one position estimation target 2-j among the moving or stopped position estimation targets 2-1 to 2-P by inputting the input feature amount into the position estimation model, and outputting the estimated and calculated position information. The position estimation model utilization unit 1-3 may use a position estimation model that has been generated in advance, or may use a position estimation model that is generated and updated by the position estimation model training unit 1-4.

The location estimation model is a model generated by training the relationship between the location information of the location estimation target 2-j and the channel information (≒ input feature) related to radio wave propagation acquired from the fixed terminal 3-i using machine learning, and is a model that outputs location information using the input feature. Alternatively, the location estimation model may be generated by generating a space equivalent to the real space in a simulation space using digital twin technology, etc., and using the relationship between the virtually generated location estimation target and the channel information calculated by simulation. The location estimation model may be a location estimation model created from the relationship between the channel information measured by another location estimation unit and the location estimation target.

The location estimation model training unit 1-4 has a function of generating a location estimation model by separately acquiring data on the location information of the location estimation target 2-j and training an estimated location estimation model capable of estimating the location of the location estimation target 2-j based on the relationship between the acquired location information and channel information. The location estimation model training unit 1-4 also has a function of updating the generated location estimation model.

The location information of the location estimation target 2-j can be obtained by periodically collecting location measurement data by a location measurement function installed in the location estimation target 2-j by some means. In addition, by using a machine learning algorithm, it is possible to learn the relationship between the location information of the location estimation target 2-j and the channel information obtained from the fixed terminal 3-i. Some means for acquiring the location measurement data is realized, for example, by using a sensor, camera, radio positioning, SLAM (Simultaneous Localization And Mapping), GPS (Global Positioning System), and sensors, cameras, etc. installed in the location estimation target 2-j, and requires equipment and methods other than the location estimation system and location estimation method provided by the present invention. The location measurement data is, for example, the position and time information of the location estimation target 2-j, and is stored in the storage unit. In addition, data of input features including at least the time is generated in the input feature generation unit 1-2.

The location estimation model training unit 1-4 can train the location estimation model by learning the relationship between the stored location and time information of the location estimation target 2-j and the input feature and time information also stored in the memory unit using training data organized on the same time axis.

[Overall configuration of wireless communication system (second example)]
FIG. 2 is a diagram showing an overall configuration (second example) of a wireless communication system according to this embodiment.

The wireless communication system of the second example differs from the first example in that the position estimation target 4-k (1≦k≦Q) has a wireless communication unit 4-k-1 (1≦k≦Q). The wireless communication units 1-1 to 1-R of the position estimation device 1 not only receive wireless signals transmitted from the fixed terminal 3-i, but also receive wireless signals transmitted from the position estimation target 4-k. The position estimation model utilization unit 1-3 inputs input feature amounts generated from channel information related to wireless communication from the fixed terminal 3-i and input feature amounts generated from channel information related to wireless communication from the position estimation target 4-k to the position estimation model, and obtains the position of the position estimation target 4-k as an output.

The fixed terminals 3-1 to 3-M are equipped with wireless communication units 3-1-1 to 3-M-1. The position estimation targets 4-1 to 4-Q are equipped with wireless communication units 4-1-1 to 4-Q-1. The wireless communication units 3-1-1 to 3-M-1 and wireless communication units 4-1-1 to 4-Q-1 transmit pilot signals that are known during transmission and reception, or wireless signals that include channel information between any communication unit. The any communication unit is the wireless communication units 1-1 to 1-R provided in the position estimation device 1, or any other wireless communication unit.

The position estimation device 1 receives wireless signals from the fixed terminal 3-i and wireless signals from the position estimation target 4-k via wireless communication units 1-1 to 1-R. The position estimation device 1 acquires each piece of channel information from each received wireless signal, and inputs each piece of acquired channel information to the input feature generation unit 1-2. The input feature generation unit 1-2 converts each piece of channel information into input features suitable for input to the position estimation model, and inputs the converted input features to the position estimation model utilization unit 1-3.

Then, the position estimation model utilization unit 1-3 estimates the position of the position estimation target 4-k by inputting the input features into a position estimation model that models the relationship between the position information of the position estimation target and the input features of the channel information using machine learning.

The location estimation model of the second example differs from the location estimation model of the first example in that channel information of the location estimation target 4-k is added to the input information. The location estimation model of the second example is trained to learn the relationship between the channel information of the wireless communication unit 3-i-1 and the wireless communication unit 4-k-1 and the location information of the location estimation target 4-k at the same time and the same time zone. The training method and generation method of the location estimation model of the second example are similar to the method of the first example. The location estimation model training unit 1-4 can train the location estimation model by collecting the location information and time information of the location estimation target 4-k and learning the relationship between the collected location information and time information of the location estimation target 4-k and the channel information of the wireless communication unit 3-i-1 and the wireless communication unit 4-k-1 at the same time and the same time zone.

The wireless communication system of the second example also differs from the first example in that it further includes a location estimation target information generation unit 1-5 that generates information about the location estimation target 4-k. The location estimation target information generation unit 1-5 acquires, for example, a location estimation target ID and a wireless communication ID in wireless communication of the location estimation target 4-k. In addition, the location estimation target information generation unit 1-5 can generate a category ID classified by, for example, owner information, the form of a person, a car, a drone, etc., and communication unit information such as the antenna configuration, antenna height, number of antennas, antenna shape, terminal type, communication mode, power consumption mode, and application used of the wireless communication unit 4-k-1, as unique information for improving the detection performance of the location estimation target 4-k. The information acquired and generated by the location estimation target information generation unit 1-5 is input to the location estimation model of the location estimation model utilization unit 1-3, or input to the location estimation model training unit 1-4.

The configurations of the wireless communication systems in the first and second examples can coexist and can be configured in combination. The position estimation model utilization unit 1-3 may perform position estimation for both the position estimation target 2-j that does not have a wireless communication unit and the position estimation target 4-k that has a wireless communication unit. The position estimation model training unit 1-4 may generate and train a position estimation model that can estimate the positions of both of them.

[Configuration of position estimation model utilization unit (variation)]
3 is a diagram showing a configuration (variation) of the position estimation model utilization unit 1-3. In this configuration, the position estimation model utilization unit 1-3 includes a plurality of position estimation models.

[Configuration of location estimation model utilization unit]
The location estimation model utilization unit 1-3 includes a channel information source combination determination unit 1-3-0, a plurality of location estimation models 1-3-1-1 to 1-3-1-S, and a location information calculation unit 1-3-2.

The channel information source combination determination unit 1-3-0 has a function of presetting multiple combinations of multiple wireless communication units (wireless communication units 3-1-1 to 3-M-1 of fixed terminals, wireless communication units 4-1-1 to 4-Q-1 of target wireless communication units for position estimation) and generating input features for each combination from channel information related to the wireless communication units in the combination.

The multiple location estimation models 1-3-1-1 to 1-3-1-S correspond to the multiple combinations mentioned above, and have the function of inputting the input feature values generated by the channel information source combination determination unit 1-3-0 for each combination and outputting the respective location information.

The position information calculation unit 1-3-2 has a function of calculating the position of the position estimation target 4-k using multiple pieces of position information output from multiple position estimation models 1-3-1-1 to 1-3-1-S.

[Operation of the location estimation model utilization unit]
First, when the input feature quantity is input to the position estimation model utilization unit 1-3, the channel information source combination determination unit 1-3-0 branches the input feature quantity into several combinations based on the identification ID of the fixed terminal 3-i, which is the transmission source of the wireless signal, and the position estimation target 4-k.

For example, consider a case where channel information is collected from four fixed terminals 3-1 to 3-4 and there are four location estimation models (S=4). The location estimation model utilization unit 1-3 is preset with four combinations of arbitrary combinations of the candidates for the four fixed terminals 3-1 to 3-4, such as {3-1, 3-2, 3-3, 3-4}, {3-1, 3-2}, {3-2, 3-3}, and {3-1, 3-4}. The channel information source combination determination unit 1-3-0 then branches these four combinations by copying the input feature amount, and inputs the branched input feature amount to the four location estimation models 1-3-1-1 to 1-3-1-4. For example, the channel information source combination determination unit 1-3-0 branches the input feature of the channel information relating to the wireless signal transmitted from the fixed terminal 3-2 into three so that the input feature is assigned to {3-1, 3-2, 3-3, 3-4}, {3-1, 3-2}, and {3-2, 3-3}, which includes the fixed terminal 3-2.

The location estimation models 1-3-1-1 to 1-3-1-4 have previously learned the relationship between the channel information of the above four combinations {3-1, 3-2, 3-3, 3-4}, {3-1, 3-2}, {3-2, 3-3}, and {3-1, 3-4} and the location information of the location estimation target 4-k. When the channel information for the combination of the four fixed terminals 3-i is input, the location estimation models 1-3-1-1 to 1-3-1-4 output the location information of the location estimation target 4-k. When there are K location prediction targets and the positions of a total of K targets out of 2-i or 4-k are predicted, a location estimation model utilization unit corresponding to each of them may be provided, or a model to be used in common may be generated and the predicted positions of the K location estimation targets may be output in the location prediction model. Alternatively, the location prediction model may output multiple location information including the location estimation target to be estimated at the present time, and only the information of the location prediction target to be predicted may be selected and used.

Here, an example has been described in which channel information is obtained from wireless communication unit 3-i-1 of fixed terminal 3-i, but the combination may also include channel information obtained from a wireless signal from wireless communication unit 4-k-1 of location estimation target 4-k. A location estimation model may be generated for a combination of wireless communication units 1-1 to 1-R that receive wireless signals. Furthermore, information obtained from any of the receiving wireless communication units 1-1 to 1-R may also be combined.

Then, the location information calculation unit 1-3-2 uses the multiple pieces of location information that have been input and outputs the final location information. An example of how to calculate the final location information is described below.

[Location information calculation method (first example)]
The first calculation method is a method of selecting location information from a plurality of pieces of location information. As shown in the experimental results described later, the accuracy of location information depends on the state of the fixed terminal and the location estimation target used in the calculation of the location information. Simply put, the closer the location estimation target is to the fixed terminal 3-i, the higher the detection accuracy is, so that the communication prediction accuracy of the model selected for the fixed terminal 3-i that is close to the detection target and contributes to the estimation accuracy is likely to be high. For this reason, the location information calculation unit 1-3-2 can also determine which location estimation model output to use based on the location information of the location estimation targets 2-j and 4-k.

For example, the location information calculation unit 1-3-2 first averages the multiple pieces of input location information, converts them into provisional location information, and then uses only the output of the location estimation model that has the highest estimation accuracy among the multiple provisional location information based on the average value of the location information and the provisional location information. Which location estimation model should be selected for the provisional location information can be evaluated using training data for generating the location estimation model. After generating each location estimation model, teacher data (consisting of input features and correct location information) can be input in each combination, and it can be evaluated which location estimation model has the highest accuracy of output. During prediction, the output results of the multiple location estimation models may be used as provisional location information, and a location estimation model with higher prediction performance may be reselected. The output of the reselected location estimation model and the output of the first provisional location information may be averaged.

[Location information calculation method (second example)]
The second calculation method is a method of performing weighting calculation processing on each of a plurality of pieces of position information. When the position information of the position estimation target 2-j is three-dimensional information of (X, Y, Z) and the position information of the position estimation target 2-j outputted from the position estimation models 1-3-1-1 to 1-3-1-4 is ( _X1 , _Y1 , _Z1 ) to ( _X4 , _Y4 , _Z4 ), respectively, the position information calculation unit 1-3-2 calculates the position information ( _Xout , _Yout , _Zout ) of the position estimation target 2-j to be finally outputted by using the formula (1).

w ₁ to w ₄ are weight values. Let _{w 1} +w ₂ +w ₃ +w ₄ =1. It may also be w ₁ =w ₂ =w ₃ =w ₄ =0.25. The weight value may be calculated from the spatial positional relationship between the fixed terminal and the communication partner, such as a straight line connecting the fixed terminal that transmitted the corresponding wireless signal and the position of the communication partner, or the distance from a figure obtained by connecting the fixed terminal that transmitted the corresponding multiple wireless signals and the position of the communication partner, or may be determined by an evaluation result using training data by linear regression. Furthermore, in order to change the weight depending on the position, it may be set as a function of the position information, as follows: w ₁ (x',y') to w ₄ (x',y'), w ₁ (x',y') + w ₂ (x',y') + w ₃ (x',y') + w ₄ (x',y') = 1. Here, (x',y') can be the above-mentioned tentative position information.

The location information calculation unit 1-3-2 may use a machine learning algorithm such as a support vector machine, decision tree regression, random forest regression, or neural network to calculate the most likely (X _out , Y out , Z _out ) from (X ₁ , Y ₁ , Z ₁ ) to _{(X 4 , Y 4 , Z 4} ) input using training data.

[Operation of wireless communication system (first example)]
FIG. 4 is a diagram showing the operation (first example) of the wireless communication system.

First, the wireless communication unit 3-i-1 of the fixed terminal 3-i, the wireless communication unit 4-k-1 of the location estimation target 4-k, or both, perform wireless communication with a predetermined wireless communication device (assumed to be the wireless communication unit 1-1 of the location estimation device 1) and start transmitting a wireless signal including a pilot signal or channel information (step S1).

Next, the wireless communication unit 1-1 of the position estimation device 1 receives wireless signals transmitted from the wireless communication unit 3-i-1 of the fixed terminal 3-i and the wireless communication unit 4-k-1 of the position estimation target 4-k, and acquires channel information regarding radio wave propagation from the received wireless signals (step S2).

Next, the input feature generator 1-2 converts the acquired channel information related to radio wave propagation into an input feature suitable for input to the position estimation model (step S3). The input feature is, for example, unconverted channel information, or a numerical value obtained by performing various operations on the channel information. For example, the input feature is obtained by dividing the channel matrix or channel vector corresponding to the antenna by a constant value. The absolute value or Frobenium norm of the channel matrix or channel vector is normalized to be constant. The channel matrix or channel vector is converted into angle information. The absolute value or Frobenium norm of the channel matrix or channel vector is divided by a constant value. The channel matrix or channel vector is converted into decibel units and subtracted by a constant value. The channel information also includes the received power value at the antenna, and the signal power and signal-to-noise power cost corresponding to the wireless communication unit. In addition, the moving average is used to obtain the average value of the values obtained from the channel information in the most recent fixed time, and the ratio to that value or the value normalized by that value can be used. At least one of the time series information can be used. A specific method for calculating the input feature will be described later.

In addition, the location estimation device 1 may generate auxiliary information other than the channel information to be input to the location estimation model, and add the generated auxiliary information to the input feature (step S4). The auxiliary information is, for example, specific information for identifying the location estimation target 2-j or the location estimation target 4-k, temperature and position information of cameras and sensors, status information such as the position, speed, and settings of the estimation target device and its surrounding reflective structures, temperature, humidity, congestion status, and setting information of the wireless communication system.

Finally, the location estimation model utilization unit 1-3 inputs the input features of the converted channel information into the location estimation model, thereby estimating, calculating, and outputting the location information of the location estimation target 2-j (step S5).

[Operation of wireless communication system (second example)]
FIG. 5 is a diagram showing the operation (second example) of the wireless communication system.

Since the operations up to step S4 are the same as those in the first example, the explanation up to step S4 will be omitted and the operations from step S5 onwards will be explained.

After steps S3 and S4, the channel information source combination determination unit 1-3-0 of the position estimation model utilization unit 1-3 branches the input feature quantity of the channel information into multiple input feature quantities and inputs the multiple branched input feature quantities to multiple position estimation models 1-3-1-1 to 1-3-1-S, which then output multiple position information candidates for the position estimation target 2-j (step S5).

Finally, the location information calculation unit 1-3-2 performs averaging, filtering, linear processing, etc. on the multiple location information candidates, and further uses an integrated model to calculate the final location information of the location estimation target 2-j (step S6).

[Method of collecting channel information and calculating input features]
An example of a method for collecting channel information and a method for calculating input feature quantities will be described below.

[How to collect channel information and how to calculate input features (first example)]
In the first method, the fixed terminal 3-i or the location estimation target 4-k transmits a pilot signal that is known in transmission and reception. By transmitting a known pattern in advance, the wireless communication unit 1-r (1≦r≦R) of the location estimation device 1 can obtain a channel matrix between the antenna (number of receiving antennas: Mr) of its own wireless communication unit 1-r and the antenna (number of transmitting antennas: Ni) of the wireless communication unit 3-i-1 or wireless communication unit 4-k-1 that transmitted the pilot signal. If OFDM (orthogonal frequency division multiplexing) is used in various wireless communication systems, it is possible to obtain a channel matrix of subcarriers corresponding to multiple frequencies.

From the channel matrix H of "the number of transmitting antennas Ni × the number of receiving antennas Mr" obtained in this way, an input feature to be input to the position estimation model utilization unit 1-3 is generated. For example, when a channel matrix is obtained for a plurality of subcarriers by OFDM, the channel matrix of the η-th subcarrier is defined as H _η . As a method of converting into an input feature, first, as shown in formula (2), the channel matrix H _η is separated into a normalized channel matrix G _η normalized by a predetermined norm, and amplitude information γ _η or power information γ _η ² .

G _η can be set so that, for example, ||G _η || _F =1. ||·|| _F represents the Frobenium norm. γ _η generally has a large fluctuation range, and the value can change by 10 to the power of 5 or more. For this reason, it is possible to convert γ _η and γ _η ² into dB, specify their maximum and minimum values, and use values converted to express the range as a normalized value within a range of 0 to 1. It is also possible to use a value γ _all obtained by selecting or averaging multiple values for different frequency conditions and antenna conditions.

Alternatively, as shown in equation (3), the amplitude information γ _η may be separated for each antenna, and the norm value may be normalized to a certain value to obtain each column vector g _{1, η} to g _{Mr, η} and its amplitude values γ _{1, η} to γ _{Mr, η} .

g _{a, η} can be set as a prescribed vector so that, for example, ||g _{a, η} || _F = 1. It is also possible to use values obtained by converting _{γ a, η} or γ _{a, η} ² into dB, prescribing their maximum and minimum values, and expressing the range as a normalized value within a range of 0 to 1. It is also possible to use the value γ _{a, all} corresponding to the a-th column vector obtained by selecting or averaging multiple values for η.

Alternatively, as shown in equation (4), the amplitude information γ _η may be separated for each antenna, and row vectors g' _{1, η} to g' _{Ni, η} with their norms normalized to a certain value, and their amplitude values γ' _{1, η} to γ' _{Ni, η} may be obtained.

g' _{b, η} can be set as a prescribed vector so that, for example, ||g' _{b, η} || _F = 1. It is also possible to use values obtained by converting γ' _{b, η} or γ' _{b, η} ² into dB, prescribing their maximum and minimum values, and expressing the range as a normalized value within a range of 0 to 1. It is also possible to use the value γ' _{b, all} corresponding to the b-th column vector obtained by selecting or averaging multiple of these for η.

The channel matrix H _η , normalized channel matrix G _η , normalized vectors g _{a, η} , and normalized vectors g' _{b, η} can have the real and imaginary parts of each element as input features, or can be used as input information as imaginary numbers, or can be converted into another format such as angle information, or quantized.

Also, as described above, the power of the received signal may be obtained as γ _η ² at the stage of simply receiving something, without obtaining a channel matrix such as H from the received signal. Such information can be obtained as RSSI (Received Signal Strength Index) in many systems. In addition, any information related to power, such as RSRQ (Reference Signals Received Power) and RSRP (Reference Signal Received Quality) in LTE and 5G systems, may be used.

In addition, since power information fluctuates over time due to factors such as the humidity and temperature of the radio wave propagation environment and the temperatures of the terminal and base station, it is also effective to use moving averages or comparative information for the amplitude value information. In other words, if the value of amplitude information, power information, RSSI, etc. acquired at a certain time is γ[t], the calculated value as shown in equation (5) may be used as the power ratio information ω[t].

The value obtained may be used as is or may be expressed in dB.

Also, correlation matrices H _η H _η ^H , H _η ^{H H} _η generated using a channel matrix H _η can be used. Correlation matrices G _η G _η ^H , G _η ^H G _η generated using a normalized channel matrix G _η can be used. Matrices ΣH η , ΣG _η , ΣH _η H η ^H , ΣH _η H _{H η} , _{ΣG η} ^G _η ^H , ΣG _η ^H G _η obtained by summing or averaging the channel matrix H η , normalized channel matrix G η , and their correlation matrices H _{η H η H} , H _{η H} ^H _η , ^{G η} _{G η H} , and G _η ^H G _η for a plurality of frequencies can be used. Eigenvalues, diagonal matrices, and unitary matrices obtained by performing QR decomposition, SVD (Singular value decomposition), eigenvector decomposition ^, and the like of these matrices can be used.

Furthermore, by using the above matrices ΣH _η , ΣG _η , ΣH _η H _η ^H , ΣH _η ^H H _η , ΣG _η G _η ^H , and ΣG _η ^H G _η , the power characteristics for the direction of arrival for the communication device, obtained by the direction of arrival estimation technology, may be used as the input feature. For example, the values obtained by multiplying the vector components by (1, exp(jdθ), exp(j2dθ), ..., exp(jNdθ)) can be calculated for θ. It is also possible to generate θ at any angle interval from 0 to 2π, and use the outputs for multiple θ as the input feature. d is a predetermined constant. N is the number of elements in the vector.

In other words, the input feature generation unit 1-2 generates, as input features, values of the phase, amplitude, real component, imaginary component of one or more of the features related to the received power of a wireless signal, signal power, power ratio information obtained from the moving average of the received power and signal power, a channel matrix consisting of radio wave propagation coefficients between multiple antennas, a correlation matrix of the channel matrix, an operational matrix obtained by signal processing the channel matrix, an operational matrix obtained by signal processing the correlation matrix, an operational matrix obtained by signal processing a channel matrix or a correlation matrix corresponding to multiple frequencies, a unitary matrix obtained by linearly operating the channel matrix, a unitary matrix obtained by linearly operating the correlation matrix, a unitary matrix obtained by linearly operating the operational matrix, a diagonal matrix obtained by linearly operating the channel matrix, a diagonal matrix obtained by linearly operating the correlation matrix, a diagonal matrix obtained by linearly operating the operational matrix, a triangular matrix obtained by linearly operating the channel matrix, a triangular matrix obtained by linearly operating the correlation matrix, and a triangular matrix obtained by linearly operating the operational matrix, and generates values obtained by standardizing the range of the coefficients of one or more of these values. The input feature generating unit 1-2 stores such input feature values as time-series data, and outputs input feature values corresponding to a plurality of times from the past to the present to the position estimation model.

In a wireless communication system that uses equalization technology, it is possible to estimate the arrival time, power, and phase conditions of the path of an incoming electrical signal. Even channel information obtained in this way as a time series can be used as input features for a position estimation model by normalizing the power, converting it to frequency components, and using features and angle information extracted by existing direction-of-arrival technology.

[Method of collecting channel information and calculation of input features (second example)]
In the second method, the fixed terminal 3-i or the location estimation target 4-k communicates with a specific wireless base station, estimates channel information from a pilot signal transmitted from and received by the specific wireless base station, generates feedback information by quantizing the signal in some way, and transmits a wireless signal including the generated feedback information. The wireless communication unit 1-r of the location estimation device 1 receives the wireless signal and obtains channel information between the specific wireless base station and the fixed terminal 3-i or the location estimation target 4-k, which is included in the received wireless signal.

First, a pilot signal that is known for transmission and reception is transmitted from a specific known wireless base station. By transmitting a known pattern in advance, the wireless communication unit 3-i-1 of the fixed terminal 3-i or the wireless communication unit 4-k-1 of the location estimation target 4-k can obtain the channel matrix between its own receiving antenna (number of receiving antennas: Ni) and the antenna (number of transmitting antennas: Mt) of the specific wireless base station that transmitted the pilot signal. In the case of OFDM, which is used in various wireless communication systems, it is possible to obtain the channel matrix of subcarriers corresponding to multiple frequencies.

From the channel matrix _Hα of "the number of transmitting antennas Mt × the number of receiving antennas Ni" obtained in this manner, an input feature to be input to the position estimation model utilization unit 1-3 is generated. For example, when a channel matrix is obtained for a plurality of subcarriers by OFDM, the channel matrix of the η-th subcarrier is defined as _Hα,η . Here, when a specific wireless base station is the wireless communication unit 1-r of the position estimation device 1, the number of transmitting antennas Mt is equal to the number of receiving antennas Mr defined for the wireless communication unit 1-r in the first method. In this case, the channel matrix Hα _,η corresponds to the transposed matrix of the channel matrix _Hη .

Therefore, as a method of converting into input features, similar to the first method, the channel matrix H _α,η is separated into a normalized channel matrix G _α,η normalized by a predetermined norm and amplitude information γ _α,η or power information γ _η2 as shown in equation ( ⁶ ).

G _{α, η} can be set so that, for example, ||G _{α, η} || _F = 1. ||·|| _F represents the Frobenium norm. _{γ α, η} and γ _{α, η} ² may be converted to dB, their maximum and minimum values may be specified, and the converted values may be expressed as normalized values within a range of 0 to 1. A value γ _{α, all} obtained by selecting or averaging multiple of these may also be used. When γ _{α, all} is obtained by averaging, it may be averaged using true values, or may be averaged after being converted to dB, or the true values may be averaged in dB units.

Alternatively, as shown in equation (7), the amplitude information γα _,η may be separated for each antenna, and the norm value may be normalized to a certain value to obtain each column vector gα _,1,η to gα _,Ni,η and its amplitude value γα _,1,η to γα _,Ni,η .

g _α,a,η can be set as a prescribed vector so that, for example, ||g _α,a,η || _F = 1. It is also possible to use values obtained by converting _{γ α,a,η} or γ _α,a,η ² into dB, prescribing their maximum and minimum values, and expressing the range as a normalized value within a range of 0 to 1. It is also possible to use the value γ _α,a,all corresponding to the a-th column vector obtained by selecting or averaging multiple of these for η.

Alternatively, as shown in equation (8), the amplitude information γα _,η may be separated for each antenna, and row vectors g'α _,1,η ∼ g'α _,Mt,η with their norm values normalized to a certain value, and their amplitude values γ'α _,1,η ∼ γ'α _,Mt,η may be obtained.

g' _α,b,η can be set as a prescribed vector so that, for example, ||g' _α,b,η || _F = 1. It is also possible to use values obtained by converting γ' _α,b ,η or γ' _α,b,η ² into dB, prescribing their maximum and minimum values, and expressing the range as a normalized value within a range of 0 to 1. It is also possible to use the value γ' _α,b,all corresponding to the b-th column vector obtained by selecting or averaging multiple of these for η.

The channel matrix H _{α, η} , normalized channel matrix G _{α, η} , normalized vector g _{α, a, η} , and normalized vector g' _{α, b, η} can have the real and imaginary parts of each element as input features, or can be used as input information as imaginary numbers, or can be converted into another format such as angle information, or quantized.

Also, correlation matrices Hα _{,ηHα ,ηH and Hα} , _ηHHα,η generated using channel matrices _Hα,η can be used. Correlation matrices Gα, ^ηGα ,ηH _{and Gα} ^, ηHGα _,η generated using normalized channel matrices Gα,η can be used. Matrices ΣHα _{, η} ,ΣGα _,η , ^ΣHα ,η,ΣHα _, ^ηHα ,η,ΣHα, _ηHHα,η , ^ΣHα _{,ηHHα , η} ,ΣGα _{,ηGα ,ηH} and ^ΣGα _,ηHGα ^, _η obtained by summing or averaging the channel matrices _Hα ,η, normalized channel matrices ^Gα , _η and their correlation matrices _Hα,ηHα , _{ηH, Hα ,ηHHα,} ^{η ,} _{Gα , ηGα ,ηH and} ^Gα _, ^ηHGα _,η for a plurality of frequencies can be used. Eigenvalues, diagonal matrices, and unitary matrices obtained by performing QR decomposition, SVD, eigenvector decomposition, etc. of these matrices can be used.

As an example, we will show a case where the feedback of channel information used in the wireless LAN standard IEEE 802.11n/ac/ax is used. In this case, the right singular matrix obtained by SVD of the channel matrix H _α,η in the above example is compressed into angle information, which is quantized and generated corresponding to multiple frequencies, and then fed back.

Here, U _η is a left singular matrix of Ni×Ni, Σ _η is a diagonal matrix of Ni× _Ni with eigenvalues λ _1,η to λ Ni,η as diagonal elements, 0 is a matrix with zero elements of Ni×(Mt-Ni), (V _S,η V _N,η ) ^H is a right singular matrix of Mt×Mt, V _S,η is a vector corresponding to the eigenvalue, and V _N,η corresponds to the 0 matrix. Ni is the number of receiving antennas. Mt is the number of transmitting antennas. In the feedback, vectors and eigenvalues are generated for the number of streams to be spatially multiplexed. In the following, it is assumed that feedback is performed for the number of receiving antennas Ni, but any number less than Ni may be used. The vectors V _η of the right singular matrix Ni are fed back, and the SNR value obtained from the eigenvalues. The SNR is not calculated for each subcarrier, but is averaged. For example, λ _1,η to λ _Ni,η are averaged over the produced subcarriers, and λ _1, to λ _Ni are divided by the thermal noise level to provide feedback as S ₁ to S _Ni . V _η can be converted to angle information (φ, ψ) and fed back as shown in equation (10).

This expression is written with a focus on a certain frequency, and the V matrix in equation (10) exists for the specified number of subcarriers, and angle information is generated for each subcarrier. It is then quantized with a specified number of quantization bits, stored in a wireless signal, and transmitted. The wireless communication unit 1-r of the position estimation device 1 can obtain this angle information and SNR information, and can also obtain the RSSI information of the wireless signal.

The angle information may be used as the input feature as it is. The sine and cosine components calculated from the angle information may be used as the input feature. A matrix obtained by converting the angle information back to a right singular matrix using formula (10) may be used. After converting the angle information back to a right singular matrix, an averaging matrix obtained by averaging the right singular matrix or its correlation matrix in the frequency direction may be used. A matrix obtained by further performing signal processing such as QR decomposition on the averaging matrix may be used. For example, when converting back to a correlation matrix, V _η corresponding to the η-th subcarrier is obtained, while only a representative value of the SNR as a whole is obtained. For this reason, a pseudo-correlation matrix obtained as an approximation is calculated, for example, as shown in formula (11).

Here, Θ represents the group of subcarriers from which V _η is obtained, and N _Θ is the number of elements of Θ. Since the lower triangular part and the upper triangular part of the off-diagonal elements of the correlation matrix obtained here are complex conjugates of each other, when using them as input features in machine learning, the diagonal elements and the lower left half or upper right half of the off-diagonal elements of the pseudo-correlation matrix R obtained here can be used. Since the off-diagonal elements are imaginary numbers, general machine learning algorithms can be applied by using the real and imaginary parts as real numbers and using them as input features.

In addition, in equation (11), even if the number of columns of V _η fed back or the number of SNRs (eigenvalues) change over time, in order to prevent the correlation matrix of equation (11) from being affected, the correlation matrix of equation (11) may be calculated using only the first vector and the first SNR of V _η , or the number of columns and SNRs of vector number N _Θ ', which is smaller than the number of columns currently obtained (N _Θ in the case of equation (11)), may be selected to calculate the correlation matrix and obtain it as the input feature.

By calculating as above, the elements of the calculated R can be in the range of 0 to 1 or -1 to 1, and the standardization process can be omitted when inputting it into deep learning.

Furthermore, as mentioned above, when calculating the correlation matrix by selecting the first vector, the first diagonal element is 1, so simply

We can calculate the correlation matrix as follows: where V _η,1 represents the first column vector of V _η .

In addition, when the right singular matrix V is restored from angle information, the imaginary part of the last element of each column vector is always 0, so if the right singular matrix is obtained as an M×Ni matrix, the value 2×Mt×Ni-Ni becomes meaningful information from the numerical values of the real and imaginary parts of each element. For example, if the right singular matrix is a 4×1 matrix, a total of 7 elements with a real part of 4 and an imaginary part of 3 are meaningful information, and if a 4×2 matrix is obtained, a total of 14 elements with a real part of 8 and an imaginary part of 6 are meaningful information. Since the imaginary part of the last element of each column is 0, the last element of each column does not need to be used.

[Experimental Results]
The position estimation method and its effects according to this embodiment will be described with reference to specific examples and indoor experimental results.

Four fixed terminals 3-1 to 3-4, one location estimation target 2-1, 4-1, and a location estimation device 1 connected to two wireless communication units 1-1, 1-2 were placed in the indoor experimental environment area shown in Figure 6. The location estimation targets 2-1, 4-1 are moving in a figure-8 pattern in the center of the experimental environment area. The fixed terminals 3-1 to 3-4 communicate wirelessly with a base station (AP), which is an access point.

The fixed terminals 3-1 to 3-4 and the location estimation target 4-1 are requested by the AP to report channel information every 100 ms, and transmit channel information angle information as feedback using the channel information feedback method defined in the wireless LAN standard IEEE 802.11ac. The AP has four antennas. The fixed terminals 3-1 to 3-4, the location estimation target 2-1, and the wireless communication units 1-1 and 1-2 each have two antennas.

The two wireless communication units 1-1 and 1-2 can obtain angle information generated from the right singular matrix of the channel matrix between the fixed terminals 3-1 to 3-4 and the AP, SNR, and the RSSI value at the wireless communication units 1-1 and 1-2 of the wireless signal from the fixed terminals 3-1 to 3-4. One RSSI was obtained from each of the wireless communication units 1-1 and 1-2.

The input feature generation unit 1-2 calculates a unitary matrix from the angle information according to the above-mentioned formula, and averages the calculated unitary matrix by frequency. This results in a total of 14 components, with 8 real parts and 6 imaginary parts. In addition, 14+2+2=18 input features, including the dB values of the two pieces of SNR information normalized to be distributed in the range of 0 to 1, and the dB values of the RSSI at the receiving antennas of the wireless communication units 1-1 and 1-2 normalized to be distributed in the range of 0 to 1, are acquired for each fixed terminal at a cycle of 100 ms.

Here we explain how to generate a location estimation model.

In order to generate a position estimation model according to this embodiment, an autonomous robot, which is a position estimation target 2-1, was run for 8 hours in an indoor experimental environment area. The lines remaining in the figure of eight in FIG. 6 are the actual path of the robot, and the running data was set so that the robot would run while passing through intermediate goals set randomly in 60 cm sections so as not to follow exactly the same path. At this time, the position estimation model training unit 1-4 generates training data in which highly accurate position information of the autonomous robot obtained from the LIDAR (light detection and ranging) and tire control information equipped in the position estimation target and the above-mentioned input feature amount related to propagation are arranged in the same time series at 200 ms intervals. As mentioned above, the generation period of the input feature amount is 100 ms, but the latest information obtained in a time width divided by a 200 ms interval was selected and used to generate training data.

Then, using the generated training data for 8 hours, a location estimation model was trained using a deep neural network that uses GRU (Gated Recurent Unit) and direct connections. The learning rate was 0.0002, and the optimization algorithm used was ADAM. The GRU had 1 hidden layer and a dimension of 35, and the weights and biases were updated by backpropagation to output X and Y coordinate information within the experimental environment area using two direct connection layers with input 35 and output 35, and one direct connection layer with input 35 and output 2. The updated location estimation model that minimized the MSE using 15 minutes of data used for validation was used.

The input feature generator 1-2 generates 18 input features for each fixed terminal 3-i as described above. The location estimation model utilization unit 1-3 uses four location estimation models using only the sets of input features (18 x 1 = 18) of each of the fixed terminals 3-1 to 3-4 as Model 1A to Model 4A, and one location estimation model using all the input features (18 x 4 = 72) of the fixed terminals 3-1 to 3-4 as Model 5A. The experimental environment area was divided into six areas, Areas A to F, as shown in Figure 7.

The position of the position estimation target 2-1, which is traveling in a figure-of-eight pattern, is then estimated using each of the five position estimation models 1A to 5A, and a decision is made as to which area (A to F) the position estimation target 2-1 belongs to, based on the average value of the five estimated position information. After that, the position information calculation unit 1-3-2 selects the output of the best position estimation model for the area to which the target belongs, and the position estimation accuracy results are evaluated using this position estimation model as model 6A.

Specifically, separate from the training data mentioned above, channel information was collected for three hours, and the location estimation results generated at this time were compared with the correct location information of the target with high accuracy that had been collected in advance as described above, to evaluate the error [m]. The location information calculation unit 1-3-2 selected Model 5A for Areas A and B, Model 4A for Area C, Model 2A for Area D, Model 4A for Area E, and Model 2A for Area F.

Figure 8 shows the results of the evaluation of location estimation error. The unit is meters. All models achieved a high estimation accuracy of less than 1m, with model 5A, which collected channel information from four fixed terminals, achieving an accuracy of less than 50cm. Furthermore, model 6A, which selected a location estimation model corresponding to the appropriate input feature depending on the location, achieved an even higher location estimation accuracy than when all channel information was always used, achieving an accuracy of less than 30cm.

Furthermore, in order to consider a case where the location estimation target has a wireless communication unit, in FIG. 6, a wireless signal is transmitted from the location estimation target 4-1 at a period of 100 ms. A case where this wireless signal is further used to generate a location estimation model is described. The location estimation model generated using the input feature amount (18×2=36) generated from both channel information of the location estimation target 4-1 and the fixed terminal 3-1 is model 1B. The location estimation model generated using the input feature amount (18×2=36) generated from both channel information of the location estimation target 4-1 and the fixed terminal 3-2 is model 2B. The location estimation model generated using the input feature amount (18×2=36) generated from both channel information of the location estimation target 4-1 and the fixed terminal 3-3 is model 3B. The location estimation model generated using the input feature amount (18×2=36) generated from both channel information of the location estimation target 4-1 and the fixed terminal 3-4 is model 4B. The location estimation model generated from the input feature values (18 x 5 = 90) collected from the location estimation target 4-1 and the five fixed terminals 3-1 to 3-4 was named model 5B. Furthermore, the location estimation results generated by models 1B to 5B were averaged to determine which area the target was in, and the most suitable location estimation model was selected, which was named model 6B.

The evaluation results of the location estimation error in this case are shown in Figure 9. Model 5B was selected in Areas A and B, Model 1B in Area C, Model 3B in Area D, and Model 4B in Areas E and F. The reason why the selected model is different from the case where only fixed terminals are used is that the contribution of the channel information of the location estimation target 4-1 is basically large, so the information with a low correlation with the location estimation target 4-1 is more important. It can also be confirmed that the location estimation accuracy is improved overall by obtaining the information of the location estimation target. In this case, model 5B, which generates a single location estimation model using the channel information of all fixed terminals, does not show a significant improvement over models 1B, 3B, and 4B. On the other hand, among the multiple location estimation models, model 6, which was selected by the location information calculation unit 1-3-2, has greatly improved the location estimation accuracy, reducing the error by 20% compared to model 5B.

In addition, an additional experiment was conducted to confirm the type of input feature used and the validity period of the generated location estimation model. In this experiment, the fixed terminals 3-1 to 3-4 shown in Figure 6 were used as they were, and only the wireless communication device 1-1 was used. Furthermore, the time interval of the input feature input to the location estimation model was set to 500 ms, and the frequency of acquisition was set to 1/5. Furthermore, 15 samples (7.5 seconds) were used in the time series direction. In this case, the location estimation model was trained using a deep neural network using GRU (Gated Recurent Unit) and direct coupling using data from 4 days (27 hours) of data from the experiment conducted over 7 days. The learning rate was 0.0002, and the optimization algorithm used was ADAM. The GRU had 1 hidden layer and a dimension of 35, and the weights and biases were updated by backpropagation to output X- and Y-coordinate information within the experimental environment area using two direct coupling layers with input 35 and output 35 and one direct coupling layer with input 35 and output 2.

Here, a total of 20 pieces of information were obtained from each of the four fixed terminals as input features: the diagonal elements (4) of the correlation matrix R calculated from the feedback information using formula C, the imaginary and real numbers (12) of the lower left triangular matrix of the off-diagonal terms, the SNR (2) included in the feedback information, and the RSSI (2) at which the feedback signal was received. As there are four fixed terminals, the total number of input features is 80.

First, deep learning was used for 27 hours of training, and the accuracy of the prediction results was evaluated based on 3 hours of data on the autonomous robot's position and channel information measured during the 7-day experiment period, which was separate from the training. This confirmed a high prediction accuracy of 0.1969 [m]. Next, to confirm how long this location prediction model would be effective, a location prediction experiment was conducted using the model 7 days after the 7-day measurement. The results are shown in Figure 10. First, looking at the line marked "No update" and the estimated accuracy for the time, we can see that the prediction accuracy exceeded 6m, making it completely unusable. Model updates are important for location prediction using fixed terminals.

Next, training data was collected for 33 minutes from 10:30 that morning, and a GRU and fully connected model was fine-tuned using a learning rate of 0.0001 and the ADAM optimization algorithm, and the autonomous robot's position for that day was predicted in the same way. The fine-tuning results are shown in Figure 10. Although the estimation accuracy changes even over the course of a day, it can be seen that after tuning, the position prediction accuracy was below 1m for about 5 hours. In this way, the position estimation model training unit can maintain high estimation accuracy by periodically collecting training data and updating the position estimation model.

[Effects of this embodiment]
According to this embodiment, the position estimation device 1 includes wireless communication units 1-1 to 1-R that receive wireless signals transmitted from wireless communication units of fixed terminals and acquire channel information related to radio wave propagation from the wireless signals, an input feature generating unit 1-2 that converts the channel information into input feature information that can be input to a position estimation model, and a position estimation model utilizing unit 1-3 that inputs the input feature information into a position estimation model that models the relationship between the channel information related to radio wave propagation and the position information of a position estimation target located in the same environment as the fixed terminal by machine learning, thereby estimating and calculating the position of the position estimation target. In other words, the position of the position estimation target is estimated using channel information related to radio wave propagation included in the wireless signal of the wireless communication terminal and the position estimation model, so that the position estimation of the position estimation target can be performed by general-purpose wireless communication, and a technology that can estimate the position of the position estimation target at low cost can be provided.

[others]
The present invention is not limited to the above-described embodiment, and various modifications are possible within the scope of the present invention.

The position estimation device 1 of the present embodiment described above can be realized, for example, as shown in FIG. 11, using a general-purpose computer system including a CPU 901, a memory 902, a storage 903, a communication device 904, an input device 905, and an output device 906. The memory 902 and the storage 903 are storage devices. In the computer system, the CPU 901 executes a predetermined program loaded onto the memory 902, thereby realizing each function of the position estimation device 1.

The position estimation device 1 may be implemented in one computer. The position estimation device 1 may be implemented in multiple computers. The position estimation device 1 may be a virtual machine implemented in a computer. The program for the position estimation device 1 may be stored in a computer-readable recording medium such as a HDD, SSD, USB memory, CD, or DVD. The program for the position estimation device 1 may also be distributed via a communication network.

1: Position estimation device
1-1 to 1-R: Wireless communication unit
1-2: Input feature generation unit
1-3: Location estimation model utilization section
1-3-0: Channel information source combination judgment unit
1-3-1-1 to 1-3-1-S: Position estimation model
1-3-2: Location information calculation section
1-4: Location estimation model training section
1-5: Location estimation target information generation unit
2-1 to 2-P: Location estimation target
3-1 to 3-M: Fixed terminals
3-1-1 to 3-M-1: Wireless communication department
4-1 to 4-Q: Location estimation target
4-1-1 to 4-Q-1: Wireless communication section
901: CPU
902: Memory
903: Storage
904: Communications equipment
905: Input device
906: Output device

Claims (6)

a wireless communication unit that receives a wireless signal transmitted from the wireless communication unit of the fixed terminal and acquires first channel information related to radio wave propagation from the wireless signal;
an input feature generator that converts the first channel information into a first input feature that can be input to a first location estimation model;
a location estimation model utilization unit that estimates and calculates a first location of a second location estimation target included in the same environment as the fixed terminal by inputting the first input feature amount to the first location estimation model that models, by machine learning, a relationship between second channel information related to radio wave propagation of a wireless signal previously transmitted from a wireless communication unit of the fixed terminal and location information of a first location estimation target located in the same environment as the fixed terminal ,
The wireless communication unit of the fixed terminal includes a plurality of wireless communication units,
The position estimation model utilization unit
a channel information source combination determination unit that has a plurality of combinations of wireless communication units in which two or more wireless communication units are combined in different ways from among the plurality of wireless communication units, and that generates a second input feature amount related to third channel information related to radio wave propagation of a wireless signal transmitted from any of the plurality of wireless communication units for a plurality of combinations including the wireless communication unit among the plurality of combinations;
a plurality of different second location estimation models corresponding to the plurality of combinations, the plurality of second location estimation models each receiving the second input feature amount and each outputting a second location of the second location estimation target;
a position information calculation unit that calculates the first position of the second position estimation target by using the second positions of the second position estimation targets output from the second position estimation models;
A position estimation device comprising: The position estimation model utilization unit
2. The position estimation device according to claim 1 , wherein the first position of the second position estimation target is estimated and calculated by inputting the first input feature and a third input feature related to fourth channel information related to radio wave propagation of a wireless signal transmitted from a wireless communication unit of the first position estimation target or the second position estimation target into the first position estimation model. The input feature generation unit
3. The position estimation device according to claim 1 or 2 , wherein, as the first input feature, among feature quantities related to the received power, signal power, power ratio information obtained from a moving average of the received power or signal power of the wireless signal, a channel matrix consisting of radio wave propagation coefficients between a plurality of antennas, a correlation matrix of the channel matrix, an operational matrix obtained by signal processing the channel matrix, an operational matrix obtained by signal processing the correlation matrix, an operational matrix obtained by signal processing the channel matrix or the correlation matrix corresponding to a plurality of frequencies, a unitary matrix obtained by linearly operating the channel matrix, a unitary matrix obtained by linearly operating the correlation matrix, a unitary matrix obtained by linearly operating the operational matrix, a diagonal matrix obtained by linearly operating the channel matrix, a diagonal matrix obtained by linearly operating the correlation matrix, a diagonal matrix obtained by linearly operating the operational matrix, a triangular matrix obtained by linearly operating the channel matrix, a triangular matrix obtained by linearly operating the correlation matrix, and a triangular matrix obtained by linearly operating the operational matrix, the device generates values obtained by standardizing values of phase, amplitude, real component, and imaginary component of one or more feature quantities, and a coefficient range of the one or more values. The input feature generation unit
4. The position estimation device according to claim 3 , wherein the first input feature amount is stored as time-series data, and the first input feature amount stored from the past to the present is output to the first position estimation model. A position estimation method performed by a position estimation device,
A first step of receiving a radio signal transmitted from a radio communication unit of a fixed terminal and acquiring first channel information related to radio wave propagation from the radio signal;
a second step of converting the first channel information into first input features that can be input to a first location estimation model;
a third step of estimating and calculating a first position of a second position estimation target included in the same environment as the fixed terminal by inputting the first input feature amount into the first position estimation model that models, by machine learning, a relationship between second channel information regarding radio wave propagation of a wireless signal previously transmitted from a wireless communication unit of the fixed terminal and position information of a first position estimation target located in the same environment as the fixed terminal ;
The wireless communication unit of the fixed terminal includes a plurality of wireless communication units,
In the third step,
a channel information source combination determination unit has a plurality of combinations of wireless communication units obtained by combining two or more wireless communication units from among the plurality of wireless communication units in different combinations, and generates a second input feature amount related to third channel information related to radio wave propagation of a wireless signal transmitted from any one of the plurality of wireless communication units for a plurality of combinations including the wireless communication unit among the plurality of combinations;
a plurality of different second location estimation models corresponding to the plurality of combinations receive the second input feature amount, and output the second location of the second location estimation target, respectively;
a position information calculation unit calculates the first position of the second position estimation target by using the second positions of the second position estimation targets output from the second position estimation models;
Location estimation method. A position estimation program that causes a computer to function as the position estimation device according to any one of claims 1 to 4 .

Images