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

Abstract

This disclosure relates to a propagation environment estimation system capable of creating an environment for increasing, through learning, the simulation accuracy for propagation characteristics of wireless radio waves. This propagation environment estimation system comprises a simulation circuit and an error calculation circuit. The simulation circuit receives data of a verification area and a wireless radio wave source. Further, the propagation characteristics of the wireless radio waves in the verification area are simulated on the basis of the data. The error calculation circuit calculates an error between the simulation results and the results of verification of the propagation characteristics of the wireless radio waves in the verification area. Further, a parameter used by a learning model of the simulation circuit is updated so as to reduce the error. The simulation circuit receives the trained learning model. Further, the propagation characteristics of the wireless radio waves are simulated by using the trained learning model for newly inputted data.

Description

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

[0001] The present disclosure relates to a propagation environment estimation system, a propagation environment estimation device, a propagation environment estimation method, and a propagation environment estimation program.

[0002] Patent Literature 1 discloses a method for estimating the propagation environment of radio waves using a scale model, which is a replica of a target area. In this method, a light source, which represents a radio wave transmitting station, and a light receiving element, which represents a radio wave receiver, are installed on the scale model. Furthermore, light irradiated from the light source is received by the light receiving element. By converting the intensity of the received light into the received intensity of the radio waves, the received intensity of the radio waves propagating through the target area can be estimated.

[0003] International Publication No. 2023 / 119661

[0004] On the other hand, there is also a known technique for creating a virtual object that represents a target area and simulating the propagation characteristics of radio waves in the target area, which eliminates the need to create a scale model.

[0005] However, with the above-mentioned method, it is not possible to improve the accuracy of the simulation of the propagation characteristics of radio waves through learning.

[0006] In order to solve the above-mentioned problems, the present disclosure aims to provide a propagation environment estimation system, a propagation environment estimation device, a propagation environment estimation method, and a propagation environment estimation program that can create an environment in which the accuracy of simulation of the propagation characteristics of wireless radio waves is improved through learning.

[0007] A first aspect of the present disclosure is preferably a propagation environment estimation system comprising: a simulation circuit; and an error calculation circuit, wherein the simulation circuit executes a process of accepting data on a verification area and a radio wave source; and a process of simulating the propagation characteristics of radio waves in the verification area based on the data, and the error calculation circuit is configured to execute a process of calculating an error between a result of the simulation and a verification result of the propagation characteristics of the radio waves in the verification area, and a process of updating parameters used by a learning model of the simulation circuit so as to reduce the error, and the simulation circuit is configured to execute a process of accepting the learned learning model, and a process of simulating the propagation characteristics of the radio waves using the learned learning model for the newly input data.

[0008] A second aspect is preferably a propagation environment estimation device equipped with a learning model, the propagation environment estimation device including a machine learning circuit that has completed desired learning by executing the following processes: a process of receiving data on a verification area and a radio wave source; a process of simulating the propagation characteristics of the radio waves in the verification area based on the data; a process of calculating an error between the result of the simulation and a verification result of the propagation characteristics of the radio waves in the verification area; and a process of updating parameters of the learning model so as to reduce the error.

[0009] A third aspect is preferably a propagation environment estimation method including: receiving data on a verification area and a radio wave source; simulating the propagation characteristics of radio waves in the verification area based on the data; calculating an error between the result of the simulation and a verification result of the propagation characteristics of the radio waves in the verification area; updating parameters used by a learning model so as to reduce the error; and simulating the propagation characteristics of the radio waves using the learned learning model for the newly input data.

[0010] A fourth aspect is preferably a propagation environment estimation program including a program that executes the following processes: a process of receiving data on a verification area and a radio wave source; a process of simulating the propagation characteristics of radio waves in the verification area based on the data; a process of calculating the error between the result of the simulation and the verification result of the propagation characteristics of the radio waves in the verification area; and a process of updating parameters of a learning model used in the simulation so as to reduce the error.

[0011] In the present disclosure, in the learning phase, the parameters of the learning model are updated to reduce the error between the simulation results and the verification results of the radio wave propagation characteristics in the verification area. This allows the learning model to complete the desired learning. Meanwhile, in the estimation phase, a simulation is performed using the trained learning model. This creates an environment in which the accuracy of the simulation of the radio wave propagation characteristics can be improved through learning.

[0012] Fig. 1 is a diagram showing a propagation environment estimation system according to embodiment 1. Fig. 2 is an example of a simulation model according to embodiment 1. Fig. 3 is a diagram showing a propagation environment estimation device according to embodiment 1. Fig. 4 is a diagram showing hardware configurations of the propagation environment estimation system and the propagation environment estimation device according to embodiment 1. Fig. 5 is a flowchart explaining processing executed by a CPU of the propagation environment estimation system according to embodiment 1.

[0013] Embodiments of the present disclosure will be described with reference to the drawings. The same or corresponding components will be designated by the same reference numerals, and repeated description may be omitted.

[0014] 1 is a diagram showing a propagation environment estimation system 140 according to embodiment 1. The propagation environment estimation system 140 includes a simulation circuit 141 and an error calculation circuit 142.

[0015] <Learning Phase> Here, the processing in the learning phase will be described first. The simulation circuit 141 receives data 10 for learning about a verification area and a radio wave source installed in the verification area. The verification area is a target area or a scale model of the target area created for estimating the radio wave propagation environment. Alternatively, the verification area may be a simulation model created for simulating the radio wave propagation characteristics using the FDTD method or the like.

[0016] The training data 10 includes, for example, information on the shape of the verification area, reflectance, and frequency of the radio wave source. Based on the training data 10, the simulation circuit 141 determines the following information 2 necessary for creating the simulation model 100:

[0017] (1) The size of the object 130 is determined so as to have an appropriate scale ratio for the frequency of the virtual electromagnetic wave source 110 installed in the simulation model 100 .

[0018] (2) The irradiation level of the virtual electromagnetic wave source 110 is determined, for example, according to the transmission strength of the radio wave source used in the verification area. However, the irradiation level may be determined so that the electromagnetic waves reach the simulation model 100, or may be determined by other methods.

[0019] (3) The frequency of the virtual electromagnetic wave source 110 is determined so as to match the behavior of radio waves propagating in the verification area. The frequency of the virtual electromagnetic wave source 110 may be a radio wave frequency or an optical frequency.

[0020] (4) The reflectance of the object 130 is determined according to the material of the structure in the verification area. If the virtual electromagnetic wave source 110 is a radio wave source, the reflectance is a radio wave reflectance. If the virtual electromagnetic wave source 110 is a light source, the reflectance is a light reflectance, and is determined to match the radio wave reflectance of each part of the verification area.

[0021] The simulation circuit 141 creates a simulation model 100 based on the above information 2.

[0022] An example of a simulation model 100 according to the first embodiment is shown in FIG. 2. The simulation model 100 includes an object 130 corresponding to a structure in the verification area, a virtual electromagnetic wave source 110, and a virtual detector 120. The virtual electromagnetic wave source 110 and the virtual detector 120 are installed so as to correspond to the installation positions of the radio wave source and the detector used in the verification area, respectively. It is desirable that the virtual detector 120 be a spherical object 130 so that it can receive electromagnetic waves from all solid angles.

[0023] Returning to the explanation of Fig. 1, the simulation circuit 141 simulates the propagation characteristics of radio waves in the verification area based on the created simulation model 100. The propagation characteristics include, for example, the reception strength, attenuation, and arrival time of radio waves.

[0024] Specifically, the simulation circuit 141 irradiates the created object 130 with electromagnetic waves from the virtual electromagnetic wave source 110 , thereby simulating the propagation characteristics of the electromagnetic waves in the virtual detector 120 .

[0025] In the simulation, a ray tracing simulation is first performed to calculate provisional values ​​of the propagation characteristics of the electromagnetic waves. The ray tracing method uses rays (light rays) that have information on the intensity and propagation direction according to the amplitude. The propagation characteristics at the virtual detector 120 are found by geometrically solving the change in the light rays from the virtual electromagnetic wave source 110, taking into account the reflectivity at each part of the object 130. Since the light rays do not contain phase information and there is no need to perform wave-like calculations, it is possible to converge the calculations in a shorter time than the FDTD method, etc.

[0026] The simulation circuit 141 inputs provisional values ​​of the propagation characteristics of the electromagnetic waves obtained by the ray tracing simulation into a model for outputting a final output value 30. The model generates the final output value 30, which is the propagation characteristics of the radio waves, from the provisional values. If the virtual electromagnetic wave source 110 is a light source and the provisional value in the ray tracing simulation is the received light intensity, the model generates the final output value 30 by converting the received light intensity into the received radio wave intensity.

[0027] The error calculation circuit 142 receives the final output value 30 from the simulation circuit 141. The error calculation circuit 142 also receives the verification result 20 of the propagation characteristics of the radio waves in the verification area. The verification result 20 includes information 3 of the propagation characteristics of the radio waves in the detector used in the verification area. The error calculation circuit 142 calculates the error between the final output value 30 calculated by the simulation circuit 141 and the verification result 20 of the propagation characteristics of the radio waves.

[0028] The error may be Root Mean Squared Error (RMSE), Mean Squared Error (MSE), cross entropy error, or the like.

[0029] The error calculation circuit 142 updates the parameters used by the learning model of the simulation circuit 141 so as to reduce the calculated error. When machine learning is based on a neural network, the parameters are the weights of the neural network. Note that the machine learning method is not limited to a neural network, and may be a random forest, an SVM (Support Vector Machine), a K-nearest neighbor method, or the like.

[0030] Furthermore, the learning model with updated parameters is received by the simulation circuit 141. As a result, the learning model of the simulation circuit 141 is in a state where the desired learning has been completed.

[0031] <Variation 1 in the Learning Phase> The propagation characteristics of the radio waves verified in the verification area do not necessarily have to be verified using a radio wave source, but may be verified using a light source. For example, in a scale model, the received light intensity of light irradiated from a light source representing a radio wave transmitting station may be converted into the received radio wave intensity, thereby obtaining the verified propagation characteristics.

[0032] <Variation 2 in the Learning Phase> In generating the final output value 30, it is not necessary to calculate provisional values ​​of the propagation characteristics of the electromagnetic waves by ray tracing simulation and then convert the provisional values ​​into the final output value 30 using a model. In other words, the final output value 30 of the propagation characteristics of the electromagnetic waves may be directly calculated based on the created simulation model 100.

[0033] <Estimation Phase> The simulation circuit 141 creates a simulation model 100 for newly input data 10, similar to the learning phase. Furthermore, the simulation circuit 141 simulates the propagation characteristics of radio waves using the trained learning model. Specifically, similar to the learning phase, the simulation circuit 141 performs a ray tracing simulation to calculate provisional values ​​of the propagation characteristics of the electromagnetic waves. Furthermore, the simulation circuit 141 inputs the calculated provisional values ​​to a model for outputting a final output value 30. The model generates the final output value 30, which is the propagation characteristics of radio waves, from the provisional values ​​using the trained learning model.

[0034] <Modification in Estimation Phase> In generating the final output value 30, as in the learning phase, it is not necessary to calculate provisional values ​​of the propagation characteristics of the electromagnetic waves by ray tracing simulation and then use a model to convert the provisional values ​​into the final output value 30. In other words, based on the created simulation model 100, the final output value 30 of the propagation characteristics of the electromagnetic waves may be directly calculated using a trained learning model.

[0035] 3 is a diagram showing a propagation environment estimation apparatus 240 according to embodiment 1. The propagation environment estimation apparatus 240 is a device that is equipped with a trained learning model. That is, the propagation environment estimation apparatus 240 includes a machine learning circuit 241 that has completed the desired learning by executing the processing described in the learning phase above.

[0036] 4 is a diagram showing the hardware configuration of the propagation environment estimation system 140 and the propagation environment estimation device 240 according to the first embodiment. The processing performed by the propagation environment estimation system 140 and the propagation environment estimation device 240 may be executed by a program using a computer having a CPU and memory and storing a propagation environment estimation program in the memory. Alternatively, the processing may be executed by a program using an integrated circuit such as an FPGA (Field Programmable Gate Array). The propagation environment estimation program may be provided by being recorded on a storage medium or via a network.

[0037] The propagation environment estimation system 140 and the propagation environment estimation device 240 have computer functions, with an input unit 40, an output unit 41, a communication unit 42, a CPU (Central Processing Unit, also called a processor) 43, a memory 44, and an HDD (Hard Disk Drive) 45 connected via a bus 46. The propagation environment estimation system 140 and the propagation environment estimation device 240 are also configured to be able to input and output data to and from a computer-readable storage medium 47.

[0038] The input unit 40 is, for example, a keyboard and a mouse, etc. The output unit 41 is, for example, a display device such as a display.

[0039] The communication unit 42 is a communication interface that communicates with, for example, a wireless device (not shown).

[0040] The memory 44 may be, for example, a volatile or non-volatile semiconductor memory such as a RAM, a ROM, or a flash memory, or a magnetic disk, a flexible disk, an optical disk, a DVD, or the like.

[0041] The CPU 43 controls each component of the propagation environment estimation system 140 and the propagation environment estimation device 240, and performs predetermined processing, etc. The memory 44 and the HDD 45 are storage devices that store data such as a propagation environment estimation program, a simulation model, a learning model, and machine learning parameters, for example.

[0042] The storage medium 47 is capable of storing a propagation environment estimation program and the like that executes the functions of the propagation environment estimation system 140 and the propagation environment estimation device 240. The storage medium 47 is a USB (Universal Serial Bus) memory, a CD-ROM (Compact Disc Read Only Memory), or the like.

[0043] The architecture of the propagation environment estimation system 140 and the propagation environment estimation device 240 is not limited to the example shown in the figure.

[0044] 5 is a flowchart illustrating processing executed by the CPU 43 of the propagation environment estimation system 140 according to embodiment 1. The CPU 43 reads a propagation environment estimation program stored in the memory 44 or the HDD 45 and executes the following processing.

[0045] First, data 10 on the verification area and the radio wave source is received (step S01).

[0046] Next, based on the data 10, an object 130 corresponding to a structure in the verification area is placed in a virtual three-dimensional space (step S02). Next, a virtual electromagnetic wave source 110 is placed on the object 130 (step S03). Next, one or more virtual detectors 120 are placed on the object 130 (step S04).

[0047] Next, the propagation characteristics of the radio waves in the virtual detector 120 are simulated (step S05). First, provisional values ​​of the propagation characteristics of the electromagnetic waves are calculated by simulation. Then, the calculated provisional values ​​are input into a model for outputting the final output value 30, thereby generating the final output value 30, which is the propagation characteristics of the radio waves.

[0048] Next, the error between the simulation result and the verification result of the radio wave propagation characteristics in the verification area is calculated (step S06). That is, the error between the final output value 30 generated in step S05 and the verification result 20 of the radio wave propagation characteristics is calculated.

[0049] Next, the machine learning parameters used by the learning model are updated so as to reduce the calculated error (step S07).

[0050] Next, newly input data 10 is accepted (step S08). Furthermore, the propagation characteristics of radio waves are simulated for the newly input data 10 using a trained learning model (step S09). Specifically, as in the learning phase, a ray tracing simulation is performed to calculate provisional values ​​of the electromagnetic wave propagation characteristics. Furthermore, the calculated provisional values ​​are input to a model for outputting a final output value 30. The model generates the final output value 30, which is the propagation characteristics of radio waves, from the provisional values ​​using the learning model that has completed training in step S07.

[0051] The processing executed by the CPU 43 of the propagation environment estimation device 240 is the processing of steps S08 to S09. The CPU 43 of the propagation environment estimation device 240 reads a propagation environment estimation program stored in the memory 44 or the HDD 45 and executes the processing of steps S08 to S09.

[0052] As described above, in the present disclosure, in the learning phase, the parameters of the learning model are updated so as to reduce the error between the simulation result and the verification result 20 of the radio wave propagation characteristics in the verification area. This brings the learning model into a state where the desired learning has been completed. On the other hand, in the estimation phase, a simulation is performed using the trained learning model. This makes it possible to provide a propagation environment estimation system, a propagation environment estimation device, a propagation environment estimation method, and a propagation environment estimation program that can create an environment in which the accuracy of the simulation of the radio wave propagation characteristics is improved through learning.

[0053] It should be noted that the present disclosure is not limited to the above-described embodiment, and various modifications can be made in the implementation stage without departing from the spirit of the present disclosure. For example, the simulation performed by the simulation circuit 141 is not limited to the ray tracing method, and any simulation based on a physical analysis theory of electromagnetic waves may be used.

[0054] The embodiments and modifications may be combined as appropriate, and in that case, the combined effects can be obtained.

[0055] 2 Information, 3 Information, 10 Data, 20 Verification result, 30 Final output value, 40 Input unit, 41 Output unit, 42 Communication unit, 43 CPU, 44 Memory, 46 Bus, 47 Storage medium, 100 Simulation model, 110 Electromagnetic wave source, 120 Detector, 130 Object, 140 Propagation environment estimation system, 141 Simulation circuit, 142 Error calculation circuit, 240 Propagation environment estimation device, 241 Machine learning circuit

Claims

1. A propagation environment estimation system comprising: a simulation circuit; and an error calculation circuit, wherein the simulation circuit executes a process of receiving data on a verification area and a radio wave source; and a process of simulating the propagation characteristics of radio waves in the verification area based on the data, and the error calculation circuit is configured to execute a process of calculating the error between the result of the simulation and the verification result of the propagation characteristics of the radio waves in the verification area, and a process of updating parameters used by a learning model of the simulation circuit so as to reduce the error, and the simulation circuit is configured to execute a process of receiving the learned learning model, and a process of simulating the propagation characteristics of the radio waves using the learned learning model for the newly input data.

2. The propagation environment estimation system of claim 1, wherein the process of simulating the propagation characteristics of the radio waves in the verification area includes: a process of calculating provisional values ​​of the propagation characteristics by simulation based on physical analysis theory; and a process of inputting the provisional values ​​into a model for outputting a final output value and generating the final output value of the propagation characteristics using the model; wherein the error calculation circuit, in calculating the error, calculates the error between the final output value and the verification result; and, in updating the parameters, updates the parameters of the learning model so as to reduce the error between the final output value and the verification result.

3. A propagation environment estimation system according to claim 1 or 2, wherein the propagation characteristics are at least one of the reception strength, attenuation, and arrival time of the radio wave.

4. A propagation environment estimation system according to claim 1 or 2, wherein the process of simulating the propagation characteristics of the radio waves simulates the propagation characteristics in a virtual detector capable of detecting electromagnetic waves from all solid angles.

5. A propagation environment estimation device equipped with a learning model, the propagation environment estimation device including a machine learning circuit that has completed the desired learning by executing the following processes: a process of receiving data on a verification area and a radio wave source; a process of simulating the propagation characteristics of radio waves in the verification area based on the data; a process of calculating the error between the result of the simulation and the verification result of the propagation characteristics of the radio waves in the verification area; and a process of updating the parameters of the learning model so as to reduce the error.

6. A propagation environment estimation method comprising: receiving data on a verification area and a radio wave source; simulating the propagation characteristics of radio waves in the verification area based on the data; calculating the error between the simulation result and the verification result of the propagation characteristics of the radio waves in the verification area; updating parameters used by a learning model so as to reduce the error; and simulating the propagation characteristics of the radio waves using the learned learning model for the newly input data.

7. A propagation environment estimation program including a program that executes the following processes: a process of receiving data on a verification area and a radio wave source; a process of simulating the propagation characteristics of radio waves in the verification area based on the data; a process of calculating the error between the result of the simulation and the verification result of the propagation characteristics of the radio waves in the verification area; and a process of updating parameters of a learning model used in the simulation so as to reduce the error.

Images