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

Abstract

This propagation characteristic estimation device estimates a propagation characteristic between a transmission station and a reception station in a wireless communications system. More specifically, the propagation characteristic estimation device extracts a first characteristic parameter by inputting, to a first convolutional neural network, a first input image which contains at least a top view image of the arrangement of structures around the reception station. In addition, the propagation characteristic estimation device extracts a second characteristic parameter by inputting, to a second convolutional neural network, a second input image which contains at least a side view image of the arrangement of structures between the transmission station and the reception station. The propagation characteristic estimation device estimates a propagation characteristic by inputting, to a fully coupled neural network, input parameters including at least the first characteristic parameter and the second characteristic parameter.

Description

[Technical Field]

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

[0002] There are known techniques for estimating propagation characteristics (propagation loss) between a transmitting station (Transmitter: Tx) and a receiving station (Receiver: Rx) in a wireless communication system. For example, Patent Document 1, Patent Document 2, Non-Patent Document 1, Non-Patent Document 2, and Non-Patent Document 3 disclose such propagation characteristics estimation techniques.

[0003] In particular, Patent Document 2, Non-Patent Document 1, and Non-Patent Document 2 disclose estimation models that estimate propagation characteristics using a convolutional neural network (CNN). More specifically, a top-view image showing the distribution of building heights around Rx is prepared. By inputting the top-view image into a CNN, feature parameters that represent the feature amounts of the top-view image are extracted. Then, by inputting the extracted feature parameters into a fully-connected neural network (FNN), propagation characteristics are estimated.

[0004] 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 (Non-Patent Document 3). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] International Publication No. 2021 / 064999 [Patent Document 2] Japanese Patent Application Publication No. 2019-122008 [Non-patent literature]

[0006] [Non-Patent Document 1] N. Kuno and Y. Takatori, “Prediction Method by Deep-Learning for Path Loss Characteristics in an Open-Square Environment,” in 2018 International Symposium on Antennas and Propagation (ISAP), pp. 443-444, 2018. [Non-patent document 2] T. Imai, K. Kitao, and M. Inomata, “Radio Propagation Prediction Model Using Convolutional Neural Networks by Deep Learning,” in 2019 13th European Conference on Antennas and Propagation (EuCAP), pp. 1-5, 2019. [Non-patent document 3] N. 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. Summary of the Invention [Problem to be solved by the invention]

[0007] As mentioned above, a method is known for estimating propagation characteristics by inputting an upward-looking image showing the distribution of building heights around the receiver into a CNN. However, this method does not sufficiently take into account over-rooftop propagation between the transmitter and receiver.

[0008] An object of the present invention is to provide a technique that can further improve the accuracy of estimating propagation characteristics in a wireless communication system. [Means for solving the problem]

[0009] A first aspect relates to a propagation characteristic estimation method for estimating propagation characteristics between a transmitting station and a receiving station in a wireless communication system. The propagation characteristic estimation method is as follows: A process of extracting first feature parameters by inputting a first input image including at least a top-view image of a layout of structures around the receiving station into a first convolutional neural network; extracting second feature parameters by inputting a second input image, the second input image including at least a side-view image of the arrangement of structures between the transmitting station and the receiving station, into a second convolutional neural network; a process of estimating propagation characteristics by inputting input parameters including at least the first feature parameter and the second feature parameter into a fully connected neural network; Includes.

[0010] A second aspect relates to a propagation characteristics estimation device that estimates propagation characteristics between a transmitting station and a receiving station in a wireless communication system. The propagation characteristic estimation device includes one or more processors. The one or more processors A process of extracting first feature parameters by inputting a first input image including at least a top-view image of a layout of structures around the receiving station into a first convolutional neural network; extracting second feature parameters by inputting a second input image, the second input image including at least a side-view image of the arrangement of structures between the transmitting station and the receiving station, into a second convolutional neural network; a process of estimating propagation characteristics by inputting input parameters including at least the first feature parameter and the second feature parameter into a fully connected neural network; is configured to execute

[0011] A third aspect relates to a propagation characteristics estimation program for estimating propagation characteristics between a transmitting station and a receiving station in a wireless communication system. The propagation characteristic estimation program is A process of extracting first feature parameters by inputting a first input image including at least a top-view image of a layout of structures around the receiving station into a first convolutional neural network; extracting second feature parameters by inputting a second input image, the second input image including at least a side-view image of the arrangement of structures between the transmitting station and the receiving station, into a second convolutional neural network; a process of estimating propagation characteristics by inputting input parameters including at least the first feature parameter and the second feature parameter into a fully connected neural network; to be executed by the computer. [Effects of the Invention]

[0012] According to the present invention, propagation characteristics are estimated based on not only top-view images but also side-view images. This makes it possible to estimate propagation characteristics taking into account over-rooftop propagation between a transmitting station and a receiving station. As a result, the accuracy of estimating propagation characteristics is improved. Furthermore, the convergence speed of the estimation is also improved. [Brief explanation of the drawings]

[0013] [Figure 1] FIG. 10 is a conceptual diagram for explaining an outline of propagation characteristic estimation according to a comparative example. [Figure 2] FIG. 1 is a block diagram illustrating an example of a functional configuration of a propagation characteristics estimation device according to a comparative example. [Figure 3] FIG. 10 is a diagram showing an example of an upper view image included in a first input image according to a comparative example. [Figure 4] FIG. 10 is a diagram showing another example of the first input image according to the comparative example. [Figure 5] FIG. 1 is a block diagram illustrating an example of a functional configuration of a propagation characteristics estimation device according to a comparative example. [Figure 6] 1 is a conceptual diagram for explaining an overview of propagation characteristic estimation according to an embodiment of the present invention; [Figure 7] 1 is a block diagram illustrating an example of a functional configuration of a propagation characteristics estimation device according to an embodiment of the present invention. [Figure 8] FIG. 10 is a diagram showing an example of a side view image included in a second input image according to the embodiment of the present invention. [Figure 9] FIG. 10 is a diagram showing another example of the second input image according to the embodiment of the present invention. [Figure 10] 1 is a block diagram illustrating an example of a functional configuration of a propagation characteristics estimation device according to an embodiment of the present invention. [Figure 11] 1 is a diagram for explaining the effect of an embodiment of the present invention; [Figure 12] 1 is a block diagram showing an example of the configuration of a propagation characteristics estimation device according to an embodiment of the present invention; [Figure 13] 3 is a flowchart summarizing a process performed by a propagation characteristics estimation device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0014] An embodiment of the present invention will be described with reference to the accompanying drawings.

[0015] Consider the estimation of propagation characteristics (path loss) between a transmitter (Tx) and a receiver (Rx) in a wireless communication system. For example, the transmitter is a base station (BS) and the receiver is a mobile station (MS).

[0016] 1. Comparative Example Overview 1 is a conceptual diagram for explaining an outline of propagation characteristic estimation according to a comparative example, which corresponds to the technique disclosed in Non-Patent Document 2 mentioned above, for example.

[0017] In the comparative example, propagation characteristics are estimated using an estimation model based on a convolutional neural network (CNN). As is well known, CNN is a useful tool that can automatically extract features from images. CNN has a structure in which convolutional layers that locally extract features from input images and pooling layers that aggregate the extracted features are repeatedly arranged.

[0018] In the comparative example, a "top view image" of the layout of structures such as buildings is used as an input image to the CNN. The top view image is a two-dimensional image viewed from above, and can also be called a bird's-eye view image. In particular, to take into account radio wave propagation around the Rx, a top view image of the layout of structures in a predetermined range around the Rx is used. The predetermined range is, for example, a rectangular range centered on the Rx. The top view image IMG_1A may represent the height distribution of structures in a predetermined range around the Rx. Such a top view image is input to the CNN, and feature parameters representing the feature amounts of the top view image are extracted.

[0019] A fully-connected neural network (FNN) is placed after the CNN. The fully-connected neural network is trained to output a propagation loss L according to feature parameters. The feature parameters extracted by the CNN are input to the FNN, which estimates the propagation loss L. Note that system parameters related to the wireless communication system may also be input to the FNN along with the feature parameters. Examples of the system parameters include frequency, Tx antenna height, Rx antenna height, etc.

[0020] The "propagation characteristics estimation device" that performs such propagation characteristics estimation processing will be described in detail below.

[0021] 1-2. Example of functional configuration Fig. 2 is a block diagram showing an example of the functional configuration of a propagation characteristics estimation device 1 according to a comparative example. In particular, Fig. 2 shows an example of the functional configuration of the propagation characteristics estimation device 1 in the learning stage. The propagation characteristics estimation device 1 in the learning stage includes, as functional blocks, an input image generation unit 100, a model unit 200, an error calculation unit 300, and a model update unit 400.

[0022] The input image generation unit 100 includes a first input image generation unit 110. The first input image generation unit 110 generates a "first input image IMG_1" to be input to the CNN. The first input image IMG_1 includes at least an upper-view image IMG_1A.

[0023] FIG. 3 shows an example of a top-view image IMG_1A included in the first input image IMG_1. The top-view image IMG_1A is a two-dimensional image showing the arrangement of structures (e.g., buildings) in a predetermined range around the Rx, viewed from above. The predetermined range is, for example, a rectangular range. For example, the predetermined range is a rectangular range centered on the Rx. For example, the predetermined range is a rectangular range of 128 m x 128 m, and this predetermined range is divided into meshes of 2 m x 2 m each, generating a top-view image IMG_1A having a 64 x 64 mesh configuration.

[0024] In the example shown in FIG. 3, the upward-looking image IMG_1A is an image that represents the height distribution of structures in a predetermined range around Rx.

[0025] In Fig. 3, the xy plane is parallel to the horizontal plane. The xy coordinate system may be an absolute coordinate system or a relative coordinate system determined by the positional relationship between Tx and Rx. In the latter case, for example, the xy coordinate system is set so that the x-axis or y-axis is parallel to the direction of Tx as seen from Rx.

[0026] The first input image generation unit 110 automatically generates an upward-view image IMG_1A based on the map data MAP and the position data POS. The map data MAP is three-dimensional map data of the area in which the wireless communication system operates. The map data MAP includes information on the three-dimensional layout (horizontal position and height) of structures (e.g., buildings). The position data POS indicates at least the respective positions of the Tx and Rx on the map.

[0027] The first input image IMG_1 may include multiple types of images. For example, the first input image IMG_1 may include, in addition to the above-mentioned top-view image IMG_1A, a "Tx distance distribution image IMG_1B" and an "Rx distance distribution image IMG_1C" as shown in FIG. 4. The Tx distance distribution image IMG_1B is an image that represents the distribution of distances from the Tx within the same predetermined range as the top-view image IMG_1A. The Rx distance distribution image IMG_1C is an image that represents the distribution of distances from the Rx within the same predetermined range as the top-view image IMG_1A. The first input image generation unit 110 automatically generates the Tx distance distribution image IMG_1B and the Rx distance distribution image IMG_1C based on the position data POS.

[0028] The model unit 200 includes an estimation model that estimates the propagation loss L from an input image. More specifically, the model unit 200 includes a first CNN unit 210 and an FNN unit 230. The FNN unit 230 is located after the first CNN unit 210.

[0029] The first CNN unit 210 includes a first convolutional neural network (hereinafter referred to as CNN1) as part of an estimation model. The first CNN unit 210 acquires a first input image IMG_1 generated by the first input image generation unit 110. Then, the first CNN unit 210 inputs the first input image IMG_1 to CNN1, thereby extracting a first feature parameter FP_1 representing the feature amount of the first input image IMG_1.

[0030] The FNN unit 230 includes a fully connected neural network (hereinafter referred to as FNN) as a part of the estimation model. The FNN unit 230 acquires input parameters including at least a first feature parameter FP_1. The input parameters may include system parameters SP related to the wireless communication system in addition to the first feature parameter FP_1. Examples of the system parameters SP include a frequency used in the wireless communication system, a Tx antenna height, an Rx antenna height, etc. The FNN unit 230 calculates the propagation loss L by inputting the input parameters into the FNN.

[0031] The error calculation unit 300 acquires the propagation loss L output from the model unit 200. Then, the error calculation unit 300 calculates the estimation error by the model unit 200 by referring to the training and verification data LRN. The training and verification data LRN indicates propagation characteristics obtained by actual measurement or simulation. For example, the error calculation unit 300 calculates the estimation error based on the difference between the received power calculated from the estimated propagation loss L and the received power obtained from the training and verification data LRN.

[0032] The model update unit 400 updates the estimation model included in the model unit 200 until the estimation error converges to a certain level or less. For example, the model update unit 400 updates the weight parameters and bias values ​​in the estimation model.

[0033] Through the learning described above, a model unit 200 that estimates the propagation loss L from an input image is realized.

[0034] 5 shows an example of the functional configuration of the propagation characteristics estimation device 1 at the propagation characteristics estimation stage. The propagation characteristics estimation device 1 at the propagation characteristics estimation stage includes, as functional blocks, an input image generation unit 100, a model unit 200, and a result output unit 500. The input image generation unit 100 is as described above. The model unit 200 has already been trained. The trained model unit 200 acquires the input image generated by the input image generation unit 100, estimates the propagation loss L according to the input image, and outputs the estimation result. The result output unit 500 stores the data of the propagation loss L output from the model unit 200 in a storage device or presents it to a user.

[0035] 2. Estimation of propagation characteristics using both top-view and side-view images 2-1. Overview In an urban macrocell environment, when an Rx is far from a Tx in a non-line-of-sight environment, "over-rooftop propagation," which is radio wave propagation across the roofs of buildings between the Tx and Rx, may become dominant. In the comparative example described in Section 1 above, over-rooftop propagation between the Tx and Rx is not sufficiently taken into consideration. This embodiment proposes a propagation characteristic estimation technology that takes over-rooftop propagation into consideration.

[0036] 6 is a conceptual diagram for explaining an outline of propagation characteristic estimation according to this embodiment. Explanations that overlap with the comparative example shown in FIG. 1 will be omitted as appropriate.

[0037] According to this embodiment, in addition to the above-mentioned "top view image," a "side-view image" is also used as an input image to the CNN. The side-view image is a two-dimensional image of at least the arrangement of structures (e.g., buildings) between Tx and Rx viewed from the side. In other words, the side-view image is a two-dimensional image of at least the arrangement of structures on the line connecting Tx and Rx viewed from the side. The xy plane defining the side-view image is orthogonal to the horizontal plane. The x direction is the horizontal direction from Tx to Rx. The y direction is the vertical direction orthogonal to the x direction. The side-view image is, for example, an image of a predetermined rectangular range. The predetermined range may be larger than the range between Tx and Rx.

[0038] As shown in Figure 6, a CNN that receives top-view images and a CNN that receives side-view images are provided separately. Each CNN extracts feature parameters that represent the feature amounts of the top-view images and side-view images. The FNN is placed after the two CNNs. The FNN is trained to output a propagation loss L according to input parameters, including the feature parameters output from each CNN.

[0039] As described above, according to this embodiment, propagation characteristics are estimated based on not only top-view images but also side-view images. This makes it possible to estimate propagation characteristics taking into account over-roof propagation between Tx and Rx. As a result, the accuracy of estimating propagation characteristics is improved. Furthermore, the convergence speed of estimation is also improved.

[0040] 2-2. Functional configuration example Fig. 7 is a block diagram showing an example of the functional configuration of a propagation characteristics estimation device 1 according to this embodiment. In particular, Fig. 7 shows an example of the functional configuration of the propagation characteristics estimation device 1 in the learning stage. The propagation characteristics estimation device 1 in the learning stage includes, as functional blocks, an input image generation unit 100, a model unit 200, an error calculation unit 300, and a model update unit 400. Descriptions that overlap with those in the comparative example shown in Fig. 2 will be omitted as appropriate.

[0041] The input image generation unit 100 includes a first input image generation unit 110 and a second input image generation unit 120. The first input image generation unit 110 is the same as in the comparative example described above. The second input image generation unit 120 generates a "second input image IMG_2" to be input to the CNN. The second input image IMG_2 includes at least a side-view image IMG_2A.

[0042] FIG. 8 shows an example of a side view image IMG_2A included in the second input image IMG_2. The side view image IMG_2A is a two-dimensional image of at least the arrangement of structures (e.g., buildings) between Tx and Rx viewed from the side. In other words, the side view image IMG_2A is a two-dimensional image of at least the arrangement of structures on the line connecting Tx and Rx viewed from the side. The xy plane defining the side view image IMG_2A is perpendicular to the horizontal plane. The x direction is the horizontal direction from Tx to Rx. The y direction is the vertical direction perpendicular to the x direction. The side view image IMG_2A is, for example, an image of a predetermined rectangular range. The predetermined range may be larger than the range between Tx and Rx.

[0043] The second input image generation unit 120 automatically generates a side view image IMG_2A based on the map data MAP and the position data POS.

[0044] The second input image IMG_2 may include multiple types of images. For example, the second input image IMG_2 may include, in addition to the side view image IMG_2A described above, a "Tx distance distribution image IMG_2B" and an "Rx distance distribution image IMG_2C" as shown in FIG. 9. The Tx distance distribution image IMG_2B is an image that represents the distribution of distances from the Tx within the same predetermined range as the side view image IMG_2A. The Rx distance distribution image IMG_2C is an image that represents the distribution of distances from the Rx within the same predetermined range as the side view image IMG_2A. The second input image generation unit 120 automatically generates the Tx distance distribution image IMG_2B and the Rx distance distribution image IMG_2C based on the position data POS.

[0045] The model unit 200 includes an estimation model that estimates the propagation loss L from an input image. More specifically, the model unit 200 includes a first CNN unit 210, a second CNN unit 220, and an FNN unit 230. The FNN unit 230 is located after the first CNN unit 210 and the second CNN unit 220.

[0046] The first CNN unit 210 is the same as in the comparative example described above.

[0047] The second CNN unit 220 includes a second convolutional neural network (hereinafter referred to as CNN2) as part of the estimation model. The second CNN unit 220 acquires a second input image IMG_2 generated by the second input image generation unit 120. Then, the second CNN unit 220 inputs the second input image IMG_2 to CNN2, thereby extracting a second feature parameter FP_2 representing the feature amount of the second input image IMG_2.

[0048] The FNN unit 230 acquires input parameters including at least a first feature parameter FP_1 and a second feature parameter FP_2. The input parameters may further include system parameters SP related to the wireless communication system. Examples of the system parameters SP include a frequency used in the wireless communication system, a Tx antenna height, an Rx antenna height, etc. The FNN unit 230 calculates a propagation loss L by inputting the input parameters into the FNN.

[0049] The error calculation section 300 and the model update section 400 are the same as those in the comparative example described above.

[0050] 10 shows an example of the functional configuration of the propagation characteristics estimation device 1 at the propagation characteristics estimation stage. The propagation characteristics estimation device 1 at the propagation characteristics estimation stage includes, as functional blocks, an input image generation unit 100, a model unit 200, and a result output unit 500. The input image generation unit 100 is as described above. The model unit 200 has already been trained. The trained model unit 200 acquires the input image generated by the input image generation unit 100, estimates the propagation loss L according to the input image, and outputs the estimation result. The result output unit 500 stores the data of the propagation loss L output from the model unit 200 in a storage device or presents it to a user.

[0051] 2-3.Effects As described above, according to this embodiment, propagation characteristics are estimated based on not only the top-viewing image IMG_1A but also the side-viewing image IMG_2A. This makes it possible to estimate propagation characteristics taking into account over-rooftop propagation between Tx and Rx. As a result, the accuracy of estimating propagation characteristics is improved. Furthermore, the convergence speed of the estimation is also improved.

[0052] FIG. 11 is a diagram for explaining the effect. The horizontal axis represents the distance between Tx and Rx, and the vertical axis represents the estimated propagation loss L. In FIG. 11, (A) represents the estimation result for the comparative example, and (B) represents the estimation result for this embodiment. Measurement data in an urban area was used as the training and validation data. Note that frequency characteristics were not taken into consideration. As can be seen from FIG. 11, the variation in the estimation results according to this embodiment is smaller than that of the comparative example. The RMS error was 9.46 dB for the comparative example and 5.63 dB for this embodiment. As such, according to this embodiment, the estimation accuracy of the propagation characteristics is improved.

[0053] 3. Example of the configuration of a propagation characteristics estimation device 12 is a block diagram showing an example of the configuration of a propagation characteristics estimation device 1 according to this embodiment. The propagation characteristics estimation device 1 includes one or more processors 10 (hereinafter simply referred to as "processor 10"), one or more storage devices 20 (hereinafter simply referred to as "storage devices 20"), a user interface 40, and an I / O interface 50.

[0054] The processor 10 performs various types of information processing. For example, the processor 10 includes a CPU (Central Processing Unit). The storage device 20 stores various types of information required for processing by the processor 10. Examples of the storage device 20 include a volatile memory, a non-volatile memory, an HDD (Hard Disk Drive), and an SSD (Solid State Drive).

[0055] The propagation characteristics estimation program 30 is a computer program executed by the processor 10. When the processor 10 executes the propagation characteristics estimation program 30, the functions of the processor 10 (propagation characteristics estimation device 1) are realized. That is, when the processor 10 executes the propagation characteristics estimation program 30, the functional configuration of the propagation characteristics estimation device 1 shown in FIGS. 7 and 10 is realized. The propagation characteristics estimation program 30 is stored in the storage device 20. The propagation characteristics estimation program 30 may be recorded on a computer-readable recording medium. The propagation characteristics estimation program 30 may be provided to the propagation characteristics estimation device 1 via a network.

[0056] The processor 10 may be realized using hardware such as an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA).

[0057] The user interface 40 provides information to the user and accepts information input from the user. The user interface 40 includes an input device and a display device.

[0058] The I / O interface 50 is communicably connected to a database. The database stores various data necessary for the propagation characteristic estimation process. For example, the database stores map data MAP, position data POS, learning and verification data LRN, etc. The processor 10 can obtain necessary information from the database via the I / O interface 50.

[0059] Fig. 13 is a flowchart outlining the processing performed by the propagation characteristics estimation device 1 according to this embodiment. In particular, Fig. 13 shows processing related to learning of an estimation model.

[0060] In step S100, the propagation characteristics estimation device 1 acquires map data MAP and position data POS from the database.

[0061] In step S200, the propagation characteristics estimation device 1 generates a first input image IMG_1 and a second input image IMG_2 based on the map data MAP and the position data POS.

[0062] In step S300, the propagation characteristics estimation device 1 estimates propagation characteristics using an estimation model. More specifically, the propagation characteristics estimation device 1 extracts a first feature parameter FP_1 by inputting a first input image IMG_1 to CNN1, and extracts a second feature parameter FP_2 by inputting a second input image IMG_2 to CNN2. Furthermore, the propagation characteristics estimation device 1 estimates propagation characteristics by inputting input parameters including the first feature parameter FP_1 and the second feature parameter FP_2 to an FNN.

[0063] In step S400, the propagation characteristics estimation device 1 calculates an estimation error by comparing the estimated propagation characteristics with the learning and verification data LRN.

[0064] In step S500, the propagation characteristics estimation device 1 determines whether the estimation error has converged to a certain level or less. If the estimation error exceeds the certain level (step S500; No), the process proceeds to step S600.

[0065] In step S600, the propagation characteristics estimation device 1 updates the estimation model so as to reduce the estimation error, after which the process returns to step S300.

[0066] If the estimation error converges to a certain level or less (step S500; Yes), the learning of the estimation model is completed.

[0067] In the propagation characteristics estimation stage, steps S100 to S300 are similar. However, in step S300, the propagation characteristics estimation device 1 estimates the propagation characteristics using a trained estimation model. Then, the propagation characteristics estimation device 1 stores the estimation results of the propagation characteristics in the storage device 20. In addition, the propagation characteristics estimation device 1 presents the estimation results of the propagation characteristics to the user via the user interface 40. [Explanation of symbols]

[0068] 1...propagation characteristic estimation device, 10...processor, 20...storage device, 30...propagation characteristic estimation program, 40...user interface, 50...I / O interface, 100...input image generation unit, 110...first input image generation unit, 120...second input image generation unit, 200...model unit, 210...first CNN unit, 220...second CNN unit, 230...FNN unit, 300...error calculation unit, 400...model update unit, 500...result output unit, CNN1...first convolutional neural network, CNN2...second convolutional neural network, FNN...fully connected neural network, FP_1...first feature parameter, FP_2...second feature parameter, IMG_1...first input image, IMG_1A...upward-view image, IMG_1B...Tx distance distribution image, IMG_1C...Rx distance distribution image, IMG_2...second input image, IMG_2A...side view image, IMG_2B...Tx distance distribution image, IMG_2C...Rx distance distribution image, L...path loss, LRN...training and validation data, MAP...map data, POS...position data, SP...system parameters

Claims

1. A propagation characteristic estimation method for estimating propagation characteristics between a transmitting station and a receiving station in a wireless communication system, comprising: extracting first feature parameters by inputting a first input image, the first input image including at least a top-view image of a layout of structures around the receiving station, into a first convolutional neural network; extracting second feature parameters by inputting a second input image including at least a side-view image of a layout of structures between the transmitting station and the receiving station into a second convolutional neural network; a process of estimating the propagation characteristics by inputting input parameters including the first characteristic parameters, the second characteristic parameters, and system parameters related to the wireless communication system into a fully connected neural network; Contains Propagation characteristic estimation method.

2. 2. The propagation characteristic estimation method according to claim 1, The second input image further comprises: an image representing a distribution of distances from the transmitter station in the same range as the side-view image; an image representing a distribution of distances from the receiving station in the same range as the side-view image; Contains Propagation characteristic estimation method.

3. 3. The propagation characteristics estimation method according to claim 1 or 2, The first input image further comprises: an image representing a distribution of distances from the transmitting station in the same range as the top-view image; an image representing a distribution of distances from the receiving station in the same range as the upward-looking image; Contains Propagation characteristic estimation method.

4. 4. The propagation characteristics estimation method according to claim 1, further comprising: The upward-view image is an image representing the height distribution of the structures around the receiving station. Propagation characteristic estimation method.

5. 5. A propagation characteristics estimation method according to claim 1, further comprising: generating the first input image and the second input image based on map data of an area in which the wireless communication system operates; Further includes Propagation characteristic estimation method.

6. 6. A propagation characteristics estimation method according to claim 1, comprising: The system parameters include a frequency used in the wireless communication system, an antenna height of the transmitting station, and an antenna height of the receiving station. Propagation characteristic estimation method.

7. A propagation characteristics estimation device for estimating propagation characteristics between a transmitting station and a receiving station in a wireless communication system, comprising: one or more processors; the one or more processors: extracting first feature parameters by inputting a first input image, the first input image including at least a top-view image of a layout of structures around the receiving station, into a first convolutional neural network; extracting second feature parameters by inputting a second input image including at least a side-view image of a layout of structures between the transmitting station and the receiving station into a second convolutional neural network; a process of estimating the propagation characteristics by inputting input parameters including the first characteristic parameters, the second characteristic parameters, and system parameters related to the wireless communication system into a fully connected neural network; configured to run Propagation characteristics estimation device.

8. A propagation characteristics estimation program for estimating propagation characteristics between a transmitting station and a receiving station in a wireless communication system, comprising: extracting first feature parameters by inputting a first input image, the first input image including at least a top-view image of a layout of structures around the receiving station, into a first convolutional neural network; extracting second feature parameters by inputting a second input image including at least a side-view image of a layout of structures between the transmitting station and the receiving station into a second convolutional neural network; a process of estimating the propagation characteristics by inputting input parameters including the first characteristic parameters, the second characteristic parameters, and system parameters related to the wireless communication system into a fully connected neural network; Have your computer run Propagation characteristic estimation program.

Images