Propagation characteristic estimation system, propagation characteristic estimation device, propagation characteristic estimation method, and program for propagation characteristic estimation

Abstract

According to the present invention, training data including the positions and radio signal frequency information 28 of a transmission station Tx and a reception station Rx and a measured value L of a propagation characteristic generated thereunder is read. A top-view image 16 and a side-view image 18 in a predetermined range including the transmission station Tx and the reception station Rx included in the training data are generated on the basis of map data 20, 22. Training of a model (FNN part 26) progresses by using feature parameters extracted from each of the top-view image 16 and the side-view image 18, the frequency information 28, and the propagation characteristic L as training data. Estimation target data including the position and the radio signal frequency information 28 of the estimation targets, which are the transmission station Tx and the reception station Tx, is read. The top-view image 16 and the side-view image 18 for estimation are generated for the predetermined range including the transmission station Tx and the reception station Rx included in the estimation target data. The model is made to estimate the propagation characteristic on the basis of the feature parameter extracted from each of the top-view image 16 and the side-view image 18 for estimation and the frequency information 28 included in the estimation target data.

Description

Propagation characteristic estimation system, propagation characteristic estimation device, propagation characteristic estimation method, and propagation characteristic estimation program

[0001] This disclosure relates to a propagation characteristics estimation system, a propagation characteristics estimation device, a propagation characteristics estimation method, and a propagation characteristics estimation program, and in particular to a propagation characteristics estimation system, a propagation characteristics estimation device, a propagation characteristics estimation method, and a propagation characteristics estimation program that are suitable for estimating propagation characteristics between a transmitting station and a receiving station that perform wireless communication.

[0002] There are known techniques for estimating propagation characteristics (propagation loss) between a transmitter (Tx) and a receiver (Rx) in a wireless communication system. For example, Non-Patent Document 1 below discloses a method for estimating propagation characteristics using a model generated by using deep learning on a side-view image of a propagation path of a wireless signal.

[0003] Furthermore, Non-Patent Document 2 below discloses that in an urban macro-cell (UMa) environment, when a receiving station Rx moves away from a transmitting station Tx in a non-line-of-sight environment, "over-rooftop propagation," which is radio wave propagation that crosses the roofs of buildings between the transmitting station Tx and the receiving station Rx, becomes dominant. In such a UMa environment, as described in Non-Patent Document 1, the accuracy of estimating propagation characteristics can be improved by using an image capturing the propagation environment from a side view as basic data.

[0004] N. Kuno, M. Inomata, M. Sasaki and W. Yamada, “Deep learning-based path loss prediction using side-view images in an UMa environment,” 2022 16th European Conference on Antennas and Propagation (EuCAP), 2022, pp. 1-5N. Kita, W. Yamada, and A. Sato, “Path loss prediction model for the over-rooftop propagation environment of microwave band in suburban areas,” Electronics and Communications in Japan (Part I: Communications), vol. 90, no. 1, pp. 13-24, 2007.

[0005] However, the behavior of a radio signal traveling from a transmitting station Tx to a receiving station Rx may differ depending on the frequency of the radio signal. Therefore, in order to accurately grasp the propagation characteristics between the two, it is desirable to take the frequency of the radio signal into consideration.

[0006] However, in the conventional techniques disclosed in Non-Patent Document 1 and the like, the estimation of propagation characteristics is performed assuming that the frequency of the radio signal is single. Therefore, with the conventional techniques, it has been difficult to easily and accurately estimate the propagation characteristics for various radio signals with different frequencies.

[0007] In order to solve the above-mentioned problems, a first object of the present disclosure is to provide a propagation characteristics estimation system that can easily and accurately estimate propagation characteristics for various wireless signals with different frequencies.

[0008] A second object of the present disclosure is to provide a propagation characteristics estimation device that can easily and accurately estimate propagation characteristics for various radio signals with different frequencies.

[0009] A third object of the present disclosure is to provide a propagation characteristics estimation method for easily and accurately estimating propagation characteristics for various radio signals with different frequencies.

[0010] A fourth object of the present disclosure is to provide a propagation characteristics estimation program for easily and accurately estimating propagation characteristics for various wireless signals with different frequencies.

[0011] In order to achieve the above object, a first aspect is a propagation characteristics estimation system for estimating propagation characteristics of a radio signal transmitted and received between a transmitting station and a receiving station, the system comprising: a first database storing a plurality of sets of learning data including combinations of positions of the transmitting station and the receiving station, frequency information of the radio signal, and actual measurement values ​​of the propagation characteristics occurring under those positions; a second database storing map data; a third database storing estimation target data including positions of the transmitting station and the receiving station for which propagation characteristics are to be estimated and frequency information of the radio signal; and a propagation characteristics estimation device for estimating the propagation characteristics using a trained model, the propagation characteristics estimation device performing the following processes: generating, based on the map data, an overhead image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a top view; generating, based on the map data, a side image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; It is desirable that the system is configured to execute the following: a learning process that advances learning of the model using, as teacher data, feature parameters extracted from each of the overhead image and the side image, and frequency information and propagation characteristics included in the set of learning data; a process that generates, based on the map data, an estimation overhead image that captures, from a top view, a predetermined range that includes the transmitting stations and receiving stations included in the estimation target data; a process that generates, based on the map data, an estimation side image that captures, from a side view, a predetermined range that includes the transmitting stations and receiving stations included in the estimation target data; and an estimation process that causes the model to estimate propagation characteristics based on the feature parameters extracted from each of the estimation overhead image and the estimation side image, and frequency information included in the estimation target data.

[0012] A second aspect is a propagation characteristics estimation device that estimates propagation characteristics of a wireless signal transmitted and received between a transmitting station and a receiving station, comprising: a process of reading learning data including a combination of the positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measurement values ​​of the propagation characteristics occurring under those positions; a process of reading map data; a process of reading estimation target data including the positions of the transmitting station and the receiving station that are targets for estimating propagation characteristics and frequency information of the wireless signal; a process of generating, based on the map data, an overhead image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a top view; a process of generating, based on the map data, a side image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; and a process of proceeding with learning of a model for estimating propagation characteristics using, as training data, feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data; It is desirable that the system is configured to execute the following processes: a process of generating an estimation overhead image based on the map data, the estimation image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; a process of generating an estimation side image based on the map data, the estimation side image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and an estimation process of having the model estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

[0013] A third aspect is a propagation characteristics estimation method for estimating propagation characteristics of a wireless signal transmitted and received between a transmitting station and a receiving station, the method comprising: storing a plurality of sets of learning data in a first database, each set including a combination of the positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measurement values ​​of the propagation characteristics occurring under those positions; storing map data in a second database; storing estimation target data in a third database, each set including the positions of the transmitting station and the receiving station for which propagation characteristics are to be estimated, and frequency information of the wireless signal; generating, based on the map data, an overhead image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a top view; generating, based on the map data, a side image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; and using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as training data, to proceed with learning of a model for estimating the propagation characteristics; It is desirable that the method includes the steps of: generating an estimation overhead image based on the map data, the estimation image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; generating an estimation side image based on the map data, the estimation side image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and having the model estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

[0014] A fourth aspect is a propagation characteristics estimation program for causing a propagation characteristics estimation device to estimate propagation characteristics of a wireless signal transmitted and received between a transmitting station and a receiving station, the program comprising: a process of causing the propagation characteristics estimation device to read learning data including a combination of positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measurement values ​​of the propagation characteristics occurring under those positions; a process of causing the propagation characteristics estimation device to read map data; a process of causing the device to read estimation target data including positions of the transmitting station and the receiving station and frequency information of the wireless signal, which are targets for estimating propagation characteristics; a process of generating, based on the map data, an overhead image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a top view; a process of generating, based on the map data, a side image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; and a process of proceeding with learning of a model for estimating propagation characteristics using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as training data; It is desirable that the data include a computer-readable program that executes the following steps: generating an estimation overhead image based on the map data, the estimation image capturing a predetermined range from a top view including the transmitting stations and receiving stations included in the estimation target data; generating an estimation side image based on the map data, the estimation side image capturing a predetermined range from a side view including the transmitting stations and receiving stations included in the estimation target data; and causing the model to estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

[0015] According to the first to fourth aspects, it is possible to easily and accurately estimate propagation characteristics for various radio signals with different frequencies.

[0016] 2A and 2B are diagrams illustrating an overview of a system of a comparative example for comparison with the propagation characteristic estimation system according to the first embodiment of the present disclosure. FIG. 2A is a first overhead image illustrating, from a top view, the heights of buildings present on a propagation path formed between a transmitting station Tx and a receiving station Rx. FIG. 2B is a second overhead image illustrating, from a top view, the distribution of distances from the transmitting station Tx. FIG. 2C is a third overhead image illustrating, from a top view, the distribution of distances from the receiving station Rx. FIG. 3A is a first side image illustrating, from a side view, the heights of buildings present on a propagation path formed between the transmitting station Tx and the receiving station Rx. FIG. 3B is a second side image illustrating, from a side view, the distribution of distances from the transmitting station Tx. FIG. 3C is a third side image illustrating, from a side view, the distribution of distances from the receiving station Rx. FIG. 3A is a diagram illustrating an overview of a propagation characteristic estimation system according to the first embodiment of the present disclosure. FIG. 1 is a block diagram for explaining the function of the propagation characteristics estimation system according to the first embodiment of the present disclosure when learning a propagation characteristics estimation model. FIG. 2 is a flowchart for explaining the flow of processing when the propagation characteristics estimation system according to the first embodiment of the present disclosure learns a propagation characteristics estimation model. FIG. 3 is a block diagram for explaining the function of the propagation characteristics estimation system according to the first embodiment of the present disclosure when estimating propagation characteristics using a learned model. FIG. 4 is a diagram for explaining effects achieved by the propagation characteristics estimation system according to the first embodiment of the present disclosure.

[0017] [Overview of Comparative Example] Fig. 1 is a diagram for explaining an overview of a propagation characteristics estimation system of a comparative example. To facilitate understanding of the propagation characteristics estimation system according to the present disclosure, the system of the comparative example shown in Fig. 1 will first be described.

[0018] The comparative system is a system for estimating the characteristics of a propagation path formed between a transmitting station Tx and a receiving station Rx that perform wireless communication. Here, it is assumed that the transmitting station Tx and the receiving station Rx are located at a fixed distance from each other in an urban environment, such as a UMa environment.

[0019] 1, the system of the comparative example includes a convolutional neural network (CNN) unit 10. The CNN unit 10 includes a first processing unit 12 and a second processing unit 14.

[0020] A top-view image 16 showing a top view of the propagation path between the transmitting station Tx and the receiving station Rx is input to the first processing unit 12. The top-view image 16 is automatically generated by the system of the comparative example based on map data 20 that records the surrounding environments of the transmitting station Tx and the receiving station Rx.

[0021] A side image 18 showing a side view of the propagation path between the transmitting station Tx and the receiving station Rx is input to the second processing unit 14. The side image 18 is automatically generated by the system of the comparative example based on map data 22 that records the surrounding environments of the transmitting station Tx and the receiving station Rx.

[0022] FIG. 2 shows three examples of overhead images 16-1 to 16-3 automatically generated by a system of the comparative example. More specifically, FIG. 2(A) is a first overhead image 16-1 showing the heights of buildings on the propagation path from a top view. FIG. 2(B) is a second overhead image 16-2 showing the distribution of distances from a transmitting station Tx from a top view. FIG. 2(C) is a third overhead image 16-3 showing the distribution of distances from a receiving station Rx from a top view. Due to drawing limitations, the images are shown in black and white, but in reality, these are all color images, with height and distance information represented by color. A vertical scale showing the relationship between color and numerical value is shown on the right side of each image.

[0023] FIG. 3 shows examples of three types of side images 18-1 to 18-3 automatically generated by a system of the comparative example. More specifically, FIG. 3(A) is a first side image 18-1 showing the height of buildings present on the propagation path from a side view. FIG. 3(B) is a second side image 18-2 showing the distribution of distances from a transmitting station Tx from a side view. Furthermore, FIG. 3(C) is a third side image 18-3 showing the distribution of distances from a receiving station Rx from a side view. As with the case of FIG. 2, these are actually color images in which numerical information is represented by color.

[0024] The comparative example will be described again with reference to FIG. 1. The CNN unit 10 is a well-known tool for automatically extracting features from an image. In this embodiment, the CNN unit 10 has a structure in which convolutional layers that locally extract features from an input image and pooling layers that aggregate the extracted features are repeatedly arranged.

[0025] A fully-connected neural network (FNN) unit 24 is arranged downstream of the CNN unit 10. The FNN unit 24 is trained to output propagation characteristics, more specifically, propagation loss L, according to the feature parameters extracted by the CNN unit 10. In addition to the feature parameters provided by the CNN unit 10, the FNN unit 24 also receives as input scalar parameters SP representing the antenna height of the transmitting station Tx, the antenna height of the receiving station Rx, etc.

[0026] The system of the comparative example uses a plurality of training data to train the FNN unit 24. The training data includes the overhead image 16, side image 18, and scalar parameter SP described above, as well as information on the received power that reaches the receiving station Rx. The strength of the received power is a value determined by the propagation loss L superimposed on the transmitted signal. Therefore, the following description will be given assuming that the received power and the propagation loss L are synonymous.

[0027] The system of the comparative example can also use multiple pieces of verification data to perform error correction in the FNN unit 24. Like the learning model, the verification data also includes the overhead image 16, the side image 18, the scalar parameters SP, and the actual measurement value of the received power (path loss L).

[0028] After the above-described learning and error correction are completed, the system of the comparative example can accurately estimate the path loss L when given the overhead image 16, side image 18, and scalar parameters SP for the newly installed transmitting station Tx and receiving station Rx. The path loss L in a UMa environment is largely dominated by so-called over-rooftop propagation. By using the side image 18, the system of the comparative example can accurately reflect the influence of over-rooftop propagation in the FNN unit 24. Therefore, the system of the comparative example can accurately estimate the path loss L in a UMa environment.

[0029] Embodiment 1. [Configuration of Embodiment 1] Fig. 4 is a diagram for explaining an outline of a propagation characteristics estimation system of this embodiment. In Fig. 4, elements that are the same as those shown in Fig. 1 are given the same reference numerals, and their description will be simplified or omitted.

[0030] The propagation characteristics estimation system of this embodiment has the same configuration as the system of the comparative example, except that the FNN unit 24 is replaced with an FNN unit 26. The FNN unit 26 used in this embodiment uses, as a scalar parameter SP, the frequency f of the radio signal when the training data and the verification data are acquired. MHz , and the frequency f of the radio signal to be estimated MHz It has the function to accept "f MHz " is a number that expresses the frequency f of the radio signal in MHz.

[0031] In recent years, various frequencies have been used in wireless communications. Radio signals traveling from a transmitting station Tx to a receiving station Rx may exhibit different behavior depending on the frequency. The comparative system shown in FIG. 1 is based on the assumption that the frequency of the radio signal is uniform. Therefore, in the comparative system, if the frequency f of the radio signal used for learning in the FNN unit 24 differs from the frequency f of the radio signal for which the path loss L is to be estimated, the accuracy of estimating the path loss L may be reduced.

[0032] Therefore, in the first embodiment, the frequency f of the wireless signal is added to the parameters related to the estimation of the propagation loss L. Specifically, the frequency f is added to the parameters to proceed with model learning and model verification of the FNN unit 26. Furthermore, after the frequency f is added to the parameters, the propagation loss L is estimated by the trained model. As a result, according to the present embodiment, the propagation loss L can be accurately estimated for wireless signals of various frequencies f.

[0033] 5 is a block diagram for explaining the operation of the propagation characteristics estimation system 30 of this embodiment when it proceeds with model learning and model verification. As shown in the figure, the propagation characteristics estimation system 30 of this embodiment includes a first database 32 that stores learning and verification data.

[0034] The first database 32 stores a large amount of measurement data obtained by measuring the propagation loss L using three frequencies, for example, 2.2 GHz, 4.7 GHz, and 26.3 GHz. More specifically, the measurement data includes the latitude and longitude of the transmitting station Tx and the receiving station Rx, the frequency f of the radio signal, and the like. MHz , and the received power (path loss L) that reaches the receiving station Rx.

[0035] Of the information stored in the first database 32, the latitude and longitude are provided to a second database 34 that stores map data. The second database 34 compares the latitude and longitude of the transmitting station Tx and the latitude and longitude of the receiving station Rx with the map information, and outputs map data of a predetermined range surrounding the transmitting station Tx and the receiving station Rx. More specifically, the second database 34 outputs three-dimensional map data necessary for generating the overhead image 16 and the side image 18 shown in FIGS. 2 and 3.

[0036] The propagation characteristics estimation system 30 of this embodiment includes a propagation characteristics estimation device 36. The propagation characteristics estimation device 36 includes an input image generation unit 38. The map data output from the second database 34 is provided to this input image generation unit 38.

[0037] The input image generation unit 38 is a block that automatically generates input images to be provided to the CNN unit 10 shown in Fig. 4. That is, it is a block that automatically generates the three types of overhead images 16 shown in Fig. 2 and the three types of side images 18 shown in Fig. 3. The input images generated here are provided to the model unit 40.

[0038] The model unit 40 is a block that realizes the functions of the CNN unit 10 and the FNN unit 26 shown in Fig. 4. The model unit 40 receives the overhead image 16 and the side image 18 generated by the input image generation unit 38, and also receives the actual measurement values ​​of the received power (propagation loss L) corresponding to these images and the scalar parameters SP related to the actual measurements from the first database 32. The scalar parameters SP include the frequency f of the wireless signal, MHz The model unit 40 uses this input information as training data to proceed with model learning.

[0039] Of the measured data stored in the first database 32, for example, 90% selected randomly is used for model training. The remaining 10% of the measured data is used for model verification. The propagation characteristics estimation device 36 uses the training data to train the model unit 40, and then causes the model unit 40 to estimate the received power (path loss L) using the verification data.

[0040] The estimation result of the received power (propagation characteristic L) is stored in the estimation data memory 42. Then, the estimation result stored in the estimation data memory 42 and the actual measurement value of the received power (propagation loss L) included in the verification data are provided to the error calculation unit 44.

[0041] The error calculation unit 44 calculates the difference between the two as an estimation error. The propagation characteristics estimation device 36 proceeds with learning while checking the convergence of the error, and ends learning when convergence of the error is confirmed. The propagation characteristics estimation device 36 completes the model unit 40 into a trained model by proceeding with processing according to the above flow.

[0042] 6 is a flowchart illustrating the flow of processing executed in the propagation characteristics estimation device 36 to realize the above-mentioned functions. In the routine shown in FIG. 6, first, learning parameters of the model unit 40 are set (step 100). This setting is performed based on information such as the transmitting station Tx and receiving station Rx, their locations, and the frequency f of the radio signal read from the first database 32.

[0043] Next, map data for a predetermined range including the transmitting station Tx and the receiving station Rx is obtained from the second database 34 (step 102).

[0044] Next, based on the acquired map data, input images to be provided to the CNN unit 10, that is, the overhead image 16 and the side image 18 shown in FIGS. 2 and 3, are generated (step 104).

[0045] Then, the generated overhead image 16 and side image 18, and the read frequency f MHz The model calculation is carried out using the information above as an input for the model and the actual measured value of the received power read from the first database 32 as the correct answer (step 106).

[0046] Next, an error calculation is performed on the model unit 40 that has been trained through the above process (step 108). Specifically, verification data is provided to the model unit 40, and the difference between the resulting estimated value and the actual measured value included in the verification data is calculated as the error.

[0047] Next, it is determined whether the error has become small enough to satisfy the convergence condition (step 110). If it is determined that the convergence condition is not yet satisfied, the model is updated to reduce the error (step 112), and then the process of step 106 is executed again.

[0048] If it is determined in step 110 that the convergence conditions are met, it is determined that learning by the model unit 40 has been completed, and the current routine is terminated.

[0049] 7 is a block diagram for explaining the operation of the propagation characteristics estimation system 30 of this embodiment when estimating propagation characteristics in an environment to be estimated. In FIG. 7, elements similar to those shown in FIG. 5 are denoted by the same reference numerals, and their description will be simplified or omitted.

[0050] The estimation of the propagation characteristics in the environment to be estimated is performed after the learning of the model unit 40 is completed. In estimating the propagation characteristics, the positions of the transmitting station Tx and receiving station Rx to be estimated are specified, and the frequency f of the radio signal transmitted and received between them is also specified. MHz These data are assumed to be stored in the third database 46.

[0051] In the propagation characteristic estimation stage, the latitudes and longitudes of the transmitting station Tx and the receiving station Rx are provided from the third database 46 to the second database 34. Based on this information, the second database 34 provides map data of a predetermined range in the vicinity of the transmitting station Tx and the receiving station Rx to the input image generation unit 38.

[0052] The input image generation unit 38 generates an overhead image 16 and a side image 18 based on the received map data and provides them to the model unit 40. The model unit 40 also receives from a third database 46 the frequency f MHz is provided as a scalar parameter SP.

[0053] The model unit 40 receives the overhead image 16, the side image 18, and the frequency f MHz The estimated data of the received power (path loss L) is calculated by processing the above using the trained model. The calculated estimated data is output from the output unit 48 via the estimated data memory 42 that stores the data.

[0054] As described above, in this embodiment, in addition to the physical conditions in which the transmitting station Tx and the receiving station Rx are placed, the frequency f MHz The model unit 40 can be trained using the parameters f MHzAfter specifying the propagation characteristics, the model unit 40 can estimate the propagation characteristics in the environment to be estimated. Therefore, unlike the system of the comparative example, the propagation characteristics estimation system 30 of this embodiment can easily and accurately estimate the propagation characteristics for various wireless signals with different frequencies.

[0055] 5 to 7 are realized by a combination of dedicated hardware and a general computer system. The computer system includes a processing unit, various memory devices, and various interfaces. The functions of the propagation characteristics estimation device 36 are realized by the processing unit executing a propagation characteristics estimation program stored in the memory device.

[0056] 8 is a diagram showing the results of the propagation characteristics estimation by the propagation characteristics estimation system 30 of this embodiment (marked with x) in comparison with the actual measured values ​​of the propagation characteristics (marked with ·). MHz The results (marked with +) obtained when model learning and estimation were performed using only the overhead image 16 without using the above-mentioned method are also shown.

[0057] 8 shows that the estimation results (marked with x) according to this embodiment match the actual measured values ​​(marked with ·) with high accuracy. The root-mean-square error (RMS error) between the two is suppressed to 4.4 dB. The RMS error when only the overhead image 16 is used (marked with +) is 12.1 dB, so the difference between the two is clear.

[0058] [Modification of First Embodiment] Incidentally, in the first embodiment described above, the frequency f of the radio signal is input to the propagation characteristic estimation device 36 in units of MHz, but the input method is not limited to this. For example, the frequency f may be input in the following manner. - Input the frequency f in units of GHz. - Input the frequency f in units of Hz. - Divide the frequency by a predetermined maximum value _Max, normalize it to the range of 0 to 1, and input it. - Input the frequency f in logarithm, which is a format widely used in multiple regression models, etc. 10 f MHzEnter the frequency f as the free space loss value.

[0059] Alternatively, the information on the frequency f may be incorporated into the image provided to the CNN unit 10. That is, the side image 18 provided to the CNN unit 10 is composed of values ​​of 0 or 1. These values ​​may be multiplied by the frequency f to convert "1" into "f," and the image provided to the CNN unit 10 may be converted into "f."

[0060] 10 CNN unit 16 Bird's-eye view image 18 Side image 26 FNN unit 28 Frequency 30 Propagation characteristic estimation system 32 First database 34 Second database 36 Propagation characteristic estimation device 38 Input image generation unit 40 Model unit 42 Estimation data memory 44 Error calculation unit 46 Third database

Claims

1. A propagation characteristics estimation system for estimating propagation characteristics of a wireless signal transmitted between a transmitting station and a receiving station, comprising: a first database storing a plurality of sets of learning data including combinations of positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measured values ​​of the propagation characteristics occurring under those positions; a second database storing map data; a third database storing estimation target data including positions of the transmitting station and the receiving station that are targets for estimating propagation characteristics and frequency information of the wireless signal; and a propagation characteristics estimation device for estimating the propagation characteristics using a trained model, wherein the propagation characteristics estimation device performs the following steps: generating an overhead image based on the map data, the predetermined range including the transmitting station and the receiving station included in the set of learning data, viewed from above; generating a side image based on the map data, the predetermined range including the transmitting station and the receiving station included in the set of learning data, viewed from a side; and a learning process for learning the model using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as teacher data; a process of generating an estimation overhead image based on the map data, the estimation image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; a process of generating an estimation side image based on the map data, the estimation image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and an estimation process of causing the model to estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

2. A propagation characteristics estimation system as claimed in claim 1, further comprising a database for storing verification data including the same items as the learning data, wherein the propagation characteristics estimation device is configured to further execute the following processes: generating, based on the map data, a verification overhead image capturing a predetermined range including the transmitting stations and receiving stations included in the set of verification data from a top view; generating, based on the map data, a verification side image capturing a predetermined range including the transmitting stations and the receiving stations included in the set of verification data from a side view; causing the model to estimate verification propagation characteristics corresponding to feature parameters extracted from each of the verification overhead image and the verification side image and frequency information included in the set of verification data; calculating an error between the propagation characteristics included in the set of verification data and the verification propagation characteristics; and determining that learning of the model is completed when the error satisfies a convergence condition.

3. A propagation characteristic estimation system as described in claim 1, wherein the frequency information is a value normalized to the range of 0 to 1 obtained by dividing the frequency used by the radio signal by a predetermined maximum frequency.

4. The frequency information is expressed as a common logarithm log 10 The propagation characteristic estimation system according to claim 1 , wherein the value is expressed as:

5. The propagation characteristics estimation system of claim 1, wherein the frequency information is a frequency used by the wireless signal, and the learning process includes: a process of generating a replaced side image by replacing "1" included in the binary data representing the side image with the frequency, and a process of proceeding with learning of the model using feature parameters extracted from each of the overhead image and the replaced side image and the propagation characteristics included in the set of learning data as teacher data, and the estimation process includes a process of generating an estimated replaced side image by replacing "1" included in the binary data representing the estimated side image with the frequency, and a process of causing the model to estimate propagation characteristics based on the feature parameters extracted from each of the estimated overhead image and the estimated replaced side image.

6. A propagation characteristic estimation device for estimating propagation characteristics of a radio signal transmitted and received between a transmitting station and a receiving station, comprising: a process of reading learning data including a combination of the positions of the transmitting station and the receiving station, frequency information of the radio signal, and actual measured values ​​of the propagation characteristics occurring under them; a process of reading map data; a process of reading estimation target data including the positions of the transmitting station and the receiving station which are targets for estimating propagation characteristics and frequency information of the radio signal; a process of generating an overhead image based on the map data, the predetermined range including the transmitting station and the receiving station included in the set of learning data being captured from a top view; a process of generating a side image based on the map data, the predetermined range including the transmitting station and the receiving station included in the set of learning data being captured from a side view; and a process of proceeding with learning of a model for estimating propagation characteristics using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as teacher data; a process of generating, based on the map data, an estimation overhead image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; a process of generating, based on the map data, an estimation side image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and an estimation process of causing the model to estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

7. A propagation characteristic estimation method for estimating propagation characteristics of a wireless signal transmitted and received between a transmitting station and a receiving station, comprising: storing a plurality of sets of learning data in a first database, each set including a combination of the positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measured values ​​of the propagation characteristics occurring under those positions; storing map data in a second database; storing estimation target data in a third database, each set including the positions of the transmitting station and the receiving station, which are targets for estimating propagation characteristics, and frequency information of the wireless signal; generating an overhead image based on the map data, which captures a predetermined range including the transmitting station and the receiving station included in the set of learning data from a top view; generating a side image based on the map data, which captures a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; and using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as teacher data, to proceed with learning of a model for estimating the propagation characteristics; a process of generating, based on the map data, an estimation overhead image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; generating, based on the map data, an estimation side image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and having the model estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

8. A propagation characteristic estimation program for causing a propagation characteristic estimation device to estimate the propagation characteristics of a wireless signal transmitted and received between a transmitting station and a receiving station, comprising: a process for causing the propagation characteristic estimation device to load learning data including a combination of positions of the transmitting station and the receiving station, frequency information of the wireless signal, and actual measured values ​​of the propagation characteristics occurring under them; a process for loading map data; a process for loading estimation target data including positions of the transmitting station and the receiving station that are targets for estimating propagation characteristics and frequency information of the wireless signal; a process for generating, based on the map data, an overhead image capturing a predetermined range including the transmitting station and the receiving station included in a set of learning data from a top view; a process for generating, based on the map data, a side image capturing a predetermined range including the transmitting station and the receiving station included in the set of learning data from a side view; and a process for proceeding with the learning of a model for estimating propagation characteristics using feature parameters extracted from each of the overhead image and the side image, and the frequency information and propagation characteristics included in the set of learning data as teacher data; A propagation characteristic estimation program including a computer-readable program that executes the following processes: generating, based on the map data, an estimation overhead image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a top view; generating, based on the map data, an estimation side image capturing a predetermined range including the transmitting stations and receiving stations included in the estimation target data from a side view; and causing the model to estimate propagation characteristics based on feature parameters extracted from each of the estimation overhead image and the estimation side image and frequency information included in the estimation target data.

Images