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

Abstract

A propagation characteristic estimation device according to one embodiment estimates propagation characteristics of radio waves between a transmission station and a reception station. The propagation characteristic estimation device comprises: a first image generation unit that generates, as top view images, two-dimensional images respectively indicating the dispositions and heights of structures around the reception station when viewed from above; a second image generation unit that generates, as side view images, two-dimensional images respectively indicating the dispositions of side surfaces of structures between the transmission station and the reception station and the respective electric characteristics of materials constituting the side surfaces; a first CNN unit that extracts a first feature parameter by inputting each of the top view images to a convolution neural network; a second CNN unit that extracts a second feature parameter by inputting each of the side view images to the convolutional neural network; and an FNN unit that estimates the propagation characteristics by inputting the first feature parameter and the second feature parameter to a fully-connected neural network.

Description

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

[0001] The present invention relates to a technique for estimating propagation characteristics of radio waves between a transmitting station and a receiving station in a wireless communication system.

[0002] In wireless communication systems, there are known techniques for estimating propagation characteristics (propagation loss) of radio waves between a transmitter (Tx) and a receiver (Rx). For example, Patent Literature 1 and Non-Patent Literature 1 disclose estimation models for estimating propagation characteristics of radio waves by inputting feature parameters extracted using two convolutional neural networks (CNNs) into a fully-connected neural network (FNN).

[0003] In addition, in urban macrocell environments, when the Rx is far from the Tx in a non-line-of-sight environment, "over-rooftop propagation," which is radio wave propagation over the roofs of buildings between the Tx and Rx, may become dominant.

[0004] International Publication No. 2023 / 127106

[0005] 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. 1-5

[0006] However, in the past, when estimating radio wave propagation characteristics, the electrical characteristics of buildings and other structures between the transmitting and receiving stations were not taken into account, which sometimes resulted in a long time for the estimation to converge or in the inability to achieve sufficient estimation accuracy.

[0007] The present invention has been made in consideration of the above-mentioned problems, and aims to provide a propagation characteristics estimation device, a propagation characteristics estimation system, a propagation characteristics estimation method, and a propagation characteristics estimation program that can accurately estimate the propagation characteristics of radio waves.

[0008] A propagation characteristics estimation device according to one embodiment of the present invention is a propagation characteristics estimation device that estimates propagation characteristics of radio waves between a transmitting station and a receiving station that communicate wirelessly, and is characterized by having: a first image generation unit that generates, as a top-view image, a two-dimensional image that shows the layout and heights of structures around the receiving station as seen from above; a second image generation unit that generates, as a side-view image, a two-dimensional image that shows the layout of the sides of the structures between the transmitting station and the receiving station and the electrical characteristics of each of the materials that make up the sides; a first CNN unit that extracts first feature parameters by inputting each of the top-view images generated by the first image generation unit into a convolutional neural network; a second CNN unit that extracts second feature parameters by inputting each of the side-view images generated by the second image generation unit into a convolutional neural network; and an FNN unit that estimates the propagation characteristics by inputting the first feature parameters extracted by the first CNN unit and the second feature parameters extracted by the second CNN unit into a fully connected neural network.

[0009] Furthermore, a propagation characteristics estimation system according to one embodiment of the present invention is a propagation characteristics estimation system for estimating propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, the propagation characteristics estimation system including: a first image generation unit that generates, as a top-view image, a two-dimensional image that shows the layout and heights of structures around the receiving station as seen from above; a second image generation unit that generates, as a side-view image, a two-dimensional image that shows the layout of sides of structures between the transmitting station and the receiving station and the electrical characteristics of each material that makes up the sides; a first CNN unit that inputs each of the top-view images generated by the first image generation unit into a convolutional neural network to extract first feature parameters; and a second CNN unit that inputs each of the side-view images generated by the second image generation unit into a convolutional neural network to extract second feature parameters. a second CNN unit that estimates the propagation characteristics by inputting the first feature parameters extracted by the first CNN unit and the second feature parameters extracted by the second CNN unit into a fully connected neural network; a storage unit that stores verification data that is regarded as the propagation characteristics of actual radio waves between the transmitting station and the receiving station; an error calculation unit that compares the verification data stored in the storage unit with the propagation characteristics estimated by the FNN unit and calculates an error between the propagation characteristics estimated by the FNN unit and the verification data; and an update unit that updates parameters used by the first CNN unit to extract the first feature parameters and parameters used by the second CNN unit to extract the second feature parameters based on the error calculated by the error calculation unit.

[0010] Furthermore, a propagation characteristics estimation method according to one embodiment of the present invention is a propagation characteristics estimation method for estimating propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, the propagation characteristics estimation method including: a first image generation step of generating, as a top-view image, a two-dimensional image that shows the layout and heights of structures around the receiving station as seen from above; a second image generation step of generating, as a side-view image, a two-dimensional image that shows the layout of the sides of structures between the transmitting station and the receiving station and the electrical characteristics of each of the materials that make up the sides; a first CNN step of extracting first feature parameters by inputting each of the top-view images generated by the first image generation step into a convolutional neural network; a second CNN step of extracting second feature parameters by inputting each of the side-view images generated by the second image generation step into a convolutional neural network; and an FNN step of estimating the propagation characteristics by inputting the first feature parameters extracted by the first CNN step and the second feature parameters extracted by the second CNN step into a fully connected neural network.

[0011] A propagation characteristics estimation program according to one embodiment of the present invention is a propagation characteristics estimation program executed by a propagation characteristics estimation device that estimates propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, the propagation characteristics estimation program including: a first image generation unit that generates, as an upper view image, a two-dimensional image that shows the layout and height of structures around the receiving station as seen from above; a second image generation unit that generates, as a side view image, a two-dimensional image that shows the layout of sides of structures between the transmitting station and the receiving station and the electrical characteristics of each material that makes up the side; and a second image generation unit that performs convolution neural network analysis on each of the upper view images generated by the first image generation unit. a first CNN unit that extracts first feature parameters by inputting the side view images generated by the second image generation unit into a convolutional neural network; a second CNN unit that extracts second feature parameters by inputting each of the side view images generated by the second image generation unit into a convolutional neural network; and an FNN unit that estimates the propagation characteristics by inputting the first feature parameters extracted by the first CNN unit and the second feature parameters extracted by the second CNN unit into a fully connected neural network.

[0012] According to the present invention, it is possible to estimate the propagation characteristics of radio waves with high accuracy.

[0013] 1 is a diagram illustrating an example of the configuration of a propagation characteristic estimation system in a learning stage according to an embodiment; FIG. 1(a) is a diagram illustrating an example of a top-view image showing the layout and height of structures around a receiving station as viewed from above; FIG. 1(b) is a diagram illustrating an example of a top-view image showing the distance from a transmitting station as viewed from above; FIG. 1(c) is a diagram illustrating an example of a top-view image showing the distance from a receiving station as viewed from above; FIG. 1(c) is a diagram illustrating an example of an outline of a side-view image generated by a second image generation unit; FIG. 1(a) is a diagram illustrating a schematic example of LOD3 used by the second image generation unit to generate a side-view image showing the electrical characteristics of each material; FIG. 1(c) is a diagram illustrating a schematic example of LOD4 used by the second image generation unit to generate a side-view image showing the electrical characteristics of each material; FIG. 1(b) is a diagram illustrating a schematic example of data input by the second image generation unit to a CNN based on LOD4; FIG. 1(c) is a diagram illustrating a schematic example of LOD4 data (indoors) input by the second image generation unit to a CNN. 1A is a diagram illustrating an example of a side-view image showing the arrangement of the side surfaces of a structure between a transmitting station and a receiving station and the electrical characteristics of each material that makes up the side surfaces; FIG. 1B is a diagram illustrating an example of a side-view image showing the distance from the transmitting station as viewed from the side; FIG. 1C is a diagram illustrating an example of a side-view image showing the distance from the receiving station as viewed from the side; FIG. 1D is a diagram illustrating a schematic diagram of functions possessed by a model unit; FIG. 1E is a diagram illustrating an example of the configuration of a propagation characteristics estimation system in an estimation stage according to one embodiment; and FIG. 1F is a diagram illustrating an example of the hardware configuration of a propagation characteristics estimation device according to one embodiment.

[0014] A propagation characteristics estimation system according to an embodiment will be described below with reference to the drawings. The propagation characteristics estimation system according to an embodiment has a neural network that performs machine learning, and is used in both a learning stage in which radio wave propagation characteristics are learned and an estimation stage in which radio wave propagation characteristics are estimated. The propagation characteristics estimation system according to an embodiment may have the same configuration or different configurations in the learning stage and the estimation stage.

[0015] 1 is a diagram illustrating an example of the configuration of a learning-stage propagation characteristics estimation system 1 according to one embodiment. The learning-stage propagation characteristics estimation system 1 according to one embodiment includes a propagation characteristics estimation device 2 that estimates propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, for example, and storage devices (storage units) 3 and 4.

[0016] The storage device 4 is a database that stores data on structures and the like located between the transmitting station and the receiving station. For example, the storage device 4 stores BIM (LoD2, LoD3, LoD4, etc.) described below.

[0017] The storage device 3 is a database that stores verification data that is considered to be the propagation characteristics of actual radio waves between a transmitting station and a receiving station. For example, the storage device 3 stores, as verification data, measurement data (such as the received power of the receiving station) that is actually measured to measure the propagation characteristics of radio waves between a transmitting station and a receiving station, or simulation data.

[0018] The propagation characteristics estimation device 2 includes, for example, an input image generation unit 20, a model unit 22, an estimated data storage unit 24, an error calculation unit 26, and an update unit 28.

[0019] The input image generation unit 20 includes, for example, a first image generation unit 200 and a second image generation unit 202 .

[0020] The first image generation unit 200 generates a two-dimensional image showing the layout and height of structures (such as buildings) around the receiving station as viewed from above as a top-view image, and outputs the image to the model unit 22. The first image generation unit 200 may also generate a two-dimensional image showing the distance from the transmitting station as viewed from above, and a two-dimensional image showing the distance from the receiving station as viewed from above, as top-view images. The top-view image showing the distance from the transmitting station and the top-view image showing the distance from the receiving station are used to specify the position and coordinates of the structures. The top-view image can also be referred to as a bird's-eye view image.

[0021] 2A and 2B are diagrams illustrating top-view images generated by the first image generation unit 200. Fig. 2A is a diagram illustrating a top-view image showing the layout and height of structures around the receiving station as viewed from above. Fig. 2B is a diagram illustrating a top-view image showing the distance from the transmitting station as viewed from above. Fig. 2C is a diagram illustrating a top-view image showing the distance from the receiving station as viewed from above.

[0022] 2(a) to 2(c), the first image generator 200 generates, for example, three types of top-view images. The top-view images are two-dimensional images covering, for example, a rectangular area of ​​128 m x 128 m, and the rectangular area is divided into 2 m x 2 m meshes, resulting in a 64 x 64 mesh configuration.

[0023] The second image generation unit 202 ( FIG. 1 ) generates a two-dimensional image as a side-view image, which shows the arrangement of the side surfaces of structures between the transmitting station and the receiving station and the electrical properties (all or part of permittivity, conductivity, magnetic permeability, etc.) of the materials that make up the side surfaces using differences in color, shading, etc., and outputs the image to the model unit 22. For example, the second image generation unit 202 generates the side-view image using the shape, height, permittivity, conductivity, magnetic permeability, etc. of the building stored in the storage device 4.

[0024] 3 is a diagram illustrating an example of a side-view image generated by the second image generator 202. The side-view image is a two-dimensional image viewed from the side that shows the arrangement of at least the side surfaces of structures on a line connecting the transmitting station and the receiving station, and the electrical characteristics of each material that makes up the side surfaces, in the form of shapes of the structures using differences in color, shading, etc.

[0025] The xy plane defining the side-view image is perpendicular to the horizontal plane. The x direction is the horizontal direction from the transmitting station to the receiving station. The y direction is the vertical direction perpendicular to the x direction. In other words, the side-view image is, for example, an image of a predetermined rectangular area.

[0026] For example, the second image generation unit 202 generates a two-dimensional image showing the electrical characteristics of each material that makes up the side of the structure as a side view image based on information showing the appearance of the structure between the transmitting station and the receiving station.

[0027] The electrical characteristics of the side surface of a structure vary significantly depending on the material that constitutes the side surface of the structure. The electrical properties of a material (dielectric constant, conductivity, magnetic permeability, etc.) are generally values ​​that can be obtained if the material is known. Therefore, the second image generation unit 202 may use structure data created by a 3D scanner using a camera or LiDAR to identify the materials that constitute the side surface of the structure and generate a side-view image that shows the electrical characteristics of each material that constitutes the side surface of the structure.

[0028] In addition, the second image generation unit 202 may generate a two-dimensional image showing the electrical characteristics of each material that makes up the side of the structure as a side view image based on a BIM (Building Information Modeling: BIM model) of the structure between the transmitting station and the receiving station.

[0029] The second image generator 202 may generate a two-dimensional image showing the electrical characteristics of each material constituting the side of the structure as a side view image based on the Level of Details (LOD) that indicates the level of detail of the features included in the 3D city model. LOD is defined as LOD0 to LOD4 for buildings.

[0030] 4 is a diagram illustrating a schematic example of LOD3 used by the second image generator 202 to generate a side-view image showing the electrical characteristics of each material. LOD3 expresses openings such as doors and windows in addition to the building height information, exterior appearance (texture, etc.), roof and wall information reproduced by LOD2.

[0031] The data of LOD2 and LOD3 may be obtained from 3D building data (for example, Google Earth (registered trademark)) that can accurately and widely obtain information on the appearance of structures.

[0032] 5 is a diagram illustrating a schematic example of LOD4 used by the second image generator 202 to generate a side view image showing the electrical characteristics of each material. In LOD4, in addition to the information reproduced by LOD3, the interior of the building is also modeled.

[0033] 6 is a diagram illustrating a schematic example of data that the second image generation unit 202 inputs to the CNN based on LOD 4. The second image generation unit 202 uses, as input data, a cross-sectional image (a cross-sectional image viewed from the side) of a building in which the permittivity, conductivity, and magnetic permeability are indicated by different shades of gray.

[0034] 7 is a diagram illustrating a schematic example of LOD4 data (indoor) input to the CNN by the second image generation unit 202. As shown in FIG. 7, the second image generation unit 202 is configured to be able to generate side-view images showing the electrical characteristics of each material, even when, for example, a transmitting station and a receiving station are provided indoors. In other words, by applying LOD4 building data, the second image generation unit 202 can generate side-view images showing the electrical characteristics of each material from outdoors to indoors and from indoor to indoors.

[0035] The second image generating unit 202 may further generate a two-dimensional image showing the distance from the transmitting station as viewed from the side, and a two-dimensional image showing the distance from the receiving station as viewed from the side, as a side view image. The side view image showing the distance from the transmitting station and the side view image showing the distance from the receiving station are used to specify the position and coordinates of a structure. The side view image can also be referred to as a side view image.

[0036] 8A and 8B are diagrams illustrating side-view images generated by the second image generation unit 202. Fig. 8A is a diagram illustrating a side-view image showing the arrangement of the side surfaces of a structure between a transmitting station and a receiving station and the electrical characteristics of each material constituting the side surfaces. Fig. 8B is a diagram illustrating a side-view image showing the distance from the transmitting station as viewed from the side. Fig. 8C is a diagram illustrating a side-view image showing the distance from the receiving station as viewed from the side.

[0037] 8(a) to 8(c), the second image generating unit 202 generates, for example, five types of side view images. The side view image shown in Fig. 8(a) that shows the layout of the side faces of a building and the electrical properties of each of the materials that make up the side faces is, for example, three side view images that show the permittivity, conductivity, and magnetic permeability of all (or part) of the building, each in different colors and shades.

[0038] The model unit 22 (FIG. 1) includes, for example, a first CNN unit 220, a second CNN unit 222, and an FNN unit 224, and is configured with a CNN part that uses CNN and an FNN part that uses FNN. FIG. 9 is a diagram schematically illustrating the functions of the model unit 22.

[0039] The first CNN unit 220 has an internal CNN, and inputs each of the top-view images generated by the first image generation unit 200 into a convolutional neural network to extract first feature parameters and output them to the FNN unit 224.

[0040] The second CNN unit 222 has an internal CNN, and inputs each of the side view images generated by the second image generation unit 202 into a convolutional neural network to extract second feature parameters and output them to the FNN unit 224.

[0041] The FNN unit 224 has an internal FNN, and inputs the first feature parameters extracted by the first CNN unit 220 and the second feature parameters extracted by the second CNN unit 222 into a fully connected neural network to estimate the propagation characteristics (propagation loss L) of radio waves between the transmitting station and the receiving station, and outputs the estimated data to the estimated data storage unit 24.

[0042] The FNN unit 224 may input system parameters related to the wireless communication system to the FNN together with the first and second characteristic parameters. The system parameters include, for example, the frequency used by the wireless communication system, the height of the Tx antenna, the height of the Rx antenna, etc.

[0043] The estimated data storage unit 24 is a storage unit that stores the propagation characteristics of radio waves estimated by the FNN unit 224 .

[0044] The error calculation unit 26 compares the verification data stored in the memory device 3 with the propagation characteristics stored in the estimated data storage unit 24, calculates the error between the propagation characteristics estimated by the FNN unit 224 and the verification data, and outputs the error to the update unit 28.

[0045] The update unit 28 updates the parameters used by the first CNN unit 220 to extract the first feature parameter and the parameters used by the second CNN unit 222 to extract the second feature parameter, based on the error calculated by the error calculation unit 26. For example, the update unit 28 updates the parameters until the error calculated by the error calculation unit 26 converges to a predetermined level or less. The update unit 28 may also update the parameters used by the FNN unit 224.

[0046] That is, the propagation characteristics estimation system 1 trains the model unit 22 using the training data, checks the convergence of the error using the verification data, and ends the training when the error has converged.

[0047] Next, a configuration example of a propagation characteristics estimation system 1 at an estimation stage according to an embodiment will be described. Fig. 10 is a diagram showing a configuration example of the propagation characteristics estimation system 1 at an estimation stage according to an embodiment. In the configuration example of the propagation characteristics estimation system 1 at an estimation stage shown in Fig. 10, components that are substantially the same as those of the propagation characteristics estimation system 1 at a learning stage shown in Fig. 1 are assigned the same reference numerals.

[0048] The propagation characteristics estimation system 1 in the estimation stage includes a propagation characteristics estimation device 2 and a storage device 4. Here, the propagation characteristics estimation device 2 includes an input image generation unit 20, a model unit 22, an estimated data storage unit 24, and an output unit 29, and estimates the propagation characteristics (propagation loss L) of radio waves using a trained model.

[0049] The output unit 29 outputs the propagation characteristics of the radio waves stored in the estimated data storage unit 24, for example, by displaying them.

[0050] In this way, the propagation characteristics estimation system 1 according to one embodiment estimates the propagation characteristics of radio waves using an upward-view image showing the layout and height of a structure as viewed from above, and a side-view image showing the layout of the side of the structure and the electrical characteristics of each of the materials that make up the side, thereby enabling accurate estimation of the propagation characteristics of radio waves and faster convergence of the estimation.

[0051] Note that each function of the propagation characteristics estimation device 2 may be partially or entirely configured by hardware such as a programmable logic device (PLD) or a field programmable gate array (FPGA), or may be configured as a program executed by a processor such as a CPU.

[0052] For example, the propagation characteristics estimation device 2 can be realized using a computer and a program, and the program can be recorded on a storage medium or provided via a network.

[0053] Fig. 11 is a diagram illustrating an example of a hardware configuration of a propagation characteristics estimation device 2 according to an embodiment. As shown in Fig. 11, for example, the propagation characteristics estimation device 2 has an input unit 50, an output unit 51, a communication unit 52, a CPU 53, a memory 54, and an HDD 55 connected via a bus 56, and thus has the functionality of a computer. The propagation characteristics estimation device 2 is also configured to be able to input and output data to and from a computer-readable storage medium 57.

[0054] The input unit 50 is, for example, a keyboard and a mouse. The output unit 51 is, for example, a display device such as a display that outputs images. The communication unit 52 is, for example, a wired or wireless network interface, and may have a function as an output unit that outputs data to the outside.

[0055] As described above, the CPU 53 controls each component of the propagation characteristics estimation device 2 and performs predetermined processing, etc. The memory 54 and HDD 55 are storage units that store data, etc.

[0056] The storage medium 57 is capable of storing programs and the like that cause the propagation characteristics estimation apparatus 2 to execute the functions of the propagation characteristics estimation apparatus 2. Note that the architecture that constitutes the propagation characteristics estimation apparatus 2 is not limited to the example shown in FIG.

[0057] The functions performed by the components described herein may be implemented in circuitry or processing circuitry, including general purpose processors, application specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), a CPU (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to perform the described functions.

[0058] A processor includes transistors and other circuits and is considered to be circuitry or processing circuitry. A processor may be a programmed processor that executes programs stored in memory.

[0059] In this specification, a circuitry, unit, or means is hardware that is programmed to realize or performs the described functions, which may be any hardware disclosed herein or any hardware known to be programmed to realize or perform the described functions.

[0060] If the hardware is a processor considered to be a type of circuitry, the circuitry, means, or unit is a combination of the hardware and software used to configure the hardware and / or processor.

[0061] 1... Propagation characteristic estimation system, 2... Propagation characteristic estimation device, 3... Storage device, 4... Storage device, 20... Input image generation unit, 22... Model unit, 24... Estimated data storage unit, 26... Error calculation unit, 28... Update unit, 29... Output unit, 50... Input unit, 51... Output unit, 52... Communication unit, 53... CPU, 54... Memory, 55... HDD, 56... Bus, 57... Storage medium, 200... First image generation unit, 202... Second image generation unit, 220... First CNN unit, 222... Second CNN unit, 224... FNN unit

Claims

1. A propagation characteristics estimation device that estimates the propagation characteristics of radio waves between a transmitting station and a receiving station that communicate wirelessly, comprising: a first image generation unit that generates, as a top-view image, a two-dimensional image that shows the layout and heights of structures around the receiving station as seen from above; a second image generation unit that generates, as a side-view image, a two-dimensional image that shows the layout of the sides of structures between the transmitting station and the receiving station and the electrical characteristics of each of the materials that make up the sides; a first CNN unit that extracts a first feature parameter by inputting each of the top-view images generated by the first image generation unit into a convolutional neural network; a second CNN unit that extracts a second feature parameter by inputting each of the side-view images generated by the second image generation unit into a convolutional neural network; and a FNN unit that estimates the propagation characteristics by inputting the first feature parameter extracted by the first CNN unit and the second feature parameter extracted by the second CNN unit into a fully connected neural network.

2. The propagation characteristic estimation device described in claim 1, characterized in that the first image generation unit generates a two-dimensional image showing the distance from the transmitting station as viewed from above, and a two-dimensional image showing the distance from the receiving station as viewed from above, as top-view images, and the second image generation unit generates a two-dimensional image showing the distance from the transmitting station as viewed from the side, and a two-dimensional image showing the distance from the receiving station as viewed from the side, as side-view images.

3. The propagation characteristics estimation device according to claim 1 or 2, further comprising: an error calculation unit that compares verification data considered to be the propagation characteristics of actual radio waves between the transmitting station and the receiving station with the propagation characteristics estimated by the FNN unit, and calculates an error between the propagation characteristics estimated by the FNN unit and the verification data; and an update unit that updates parameters used by the first CNN unit to extract a first feature parameter and parameters used by the second CNN unit to extract a second feature parameter, based on the error calculated by the error calculation unit.

4. The propagation characteristics estimation device according to claim 1 or 2, characterized in that the second image generation unit generates, as a side-view image, a two-dimensional image showing the electrical characteristics of each material that makes up the side of the structure, based on information showing the appearance of the structure between the transmitting station and the receiving station.

5. A propagation characteristics estimation device as described in claim 1 or 2, characterized in that the second image generation unit generates a two-dimensional image showing the electrical characteristics of each material constituting the side of the structure as a side-view image based on a BIM of the structure between the transmitting station and the receiving station.

6. A propagation characteristic estimation system for estimating propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, comprising: a first image generation unit that generates, as a top-view image, a two-dimensional image that shows the layout and height of structures around the receiving station as seen from above; a second image generation unit that generates, as a side-view image, a two-dimensional image that shows the layout of the sides of structures between the transmitting station and the receiving station and the electrical characteristics of each material that makes up the sides; a first CNN unit that extracts a first feature parameter by inputting each of the top-view images generated by the first image generation unit into a convolutional neural network; a second CNN unit that extracts a second feature parameter by inputting each of the side-view images generated by the second image generation unit into a convolutional neural network; and a FNN unit that estimates the propagation characteristics by inputting the first feature parameter extracted by the first CNN unit and the second feature parameter extracted by the second CNN unit into a fully connected neural network. a memory unit that stores verification data that is regarded as a propagation characteristic of an actual radio wave between the transmitting station and the receiving station; an error calculation unit that compares the verification data stored in the memory unit with the propagation characteristic estimated by the FNN unit and calculates an error between the propagation characteristic estimated by the FNN unit and the verification data; and an update unit that updates parameters used by the first CNN unit to extract a first feature parameter and parameters used by the second CNN unit to extract a second feature parameter based on the error calculated by the error calculation unit.

7. A propagation characteristics estimation method for estimating the propagation characteristics of radio waves between a transmitting station and a receiving station that communicate wirelessly, comprising: a first image generation step of generating, as a top-view image, a two-dimensional image that shows the layout and heights of structures around the receiving station as seen from above; a second image generation step of generating, as a side-view image, a two-dimensional image that shows the layout of the sides of structures between the transmitting station and the receiving station and the electrical characteristics of each of the materials that make up the sides; a first CNN step of inputting each of the top-view images generated by the first image generation step into a convolutional neural network to extract a first feature parameter; a second CNN step of inputting each of the side-view images generated by the second image generation step into a convolutional neural network to extract a second feature parameter; and a FNN step of estimating the propagation characteristics by inputting the first feature parameter extracted by the first CNN step and the second feature parameter extracted by the second CNN step into a fully connected neural network.

8. A propagation characteristics estimation program executed by a propagation characteristics estimation device that estimates propagation characteristics of radio waves between a transmitting station and a receiving station that perform wireless communication, comprising: a first image generation unit that generates, as an upper-view image, a two-dimensional image that shows the layout and height of structures around the receiving station as viewed from above; a second image generation unit that generates, as a side-view image, a two-dimensional image that shows the layout of the sides of structures between the transmitting station and the receiving station and the electrical characteristics of each material that makes up the sides; a first CNN unit that extracts a first feature parameter by inputting each of the upper-view images generated by the first image generation unit into a convolutional neural network; a second CNN unit that extracts a second feature parameter by inputting each of the side-view images generated by the second image generation unit into a convolutional neural network; and a FNN unit that estimates the propagation characteristics by inputting the first feature parameter extracted by the first CNN unit and the second feature parameter extracted by the second CNN unit into a fully connected neural network. a propagation characteristics estimation program for causing a computer to function as each unit of the propagation characteristics estimation device having the above-mentioned steps.

Images