JP7794092B2 - Antenna design support device, learning device, and program - Google Patents

Description

The disclosed technology relates to an antenna design support device, a learning device, and a program.

Conventionally, a known antenna design support device that assists in antenna design is equipped with an antenna performance prediction unit that uses a deep-learned convolutional neural network to predict antenna performance based on a set of pixel data, in which the antenna shape is represented by multiple pixels, and performance data that indicates the antenna's performance, as training data (Patent Document 1).

WO2020/110649

Although Patent Document 1 above predicts antenna performance using a convolutional neural network, there is a demand for antenna design support devices, learning devices, and programs that offer even greater accuracy in predicting antenna performance.

The disclosed technology was developed in light of the above points, and aims to provide an antenna design support device, learning device, and program that can predict antenna performance with even greater accuracy.

A first aspect of the present disclosure is an antenna design support device that supports the design of an antenna, and includes an acquisition unit that acquires graph data representing the shape of the antenna, and a prediction unit that predicts the antenna performance of the antenna based on the graph data acquired by the acquisition unit using a pre-trained neural network, where the neural network is a graph neural network that predicts the antenna performance of the antenna based on the graph data.

A second aspect of the present disclosure is directed to the antenna design support device of the first aspect, wherein the graph data may have features including information about the shape of the antenna.

A third aspect of the present disclosure is an antenna design support device according to the second aspect, wherein the graph data is a graph including nodes representing each point, including the endpoints and intersections, of the antenna lines, and edges representing lines connecting the endpoints or intersections in a connecting relationship, and each of the nodes may have a feature including position information of each point.

A fourth aspect of the present disclosure is an antenna design support device according to the second aspect, wherein the graph data is a graph including edges representing each point, including the endpoints and intersections, of the antenna lines, and nodes representing the lines connecting the endpoints or intersections in a connecting relationship, and each of the nodes may have a feature including position information of each line.

A fifth aspect of the present disclosure relates to the antenna design support device of any one of the first to fourth aspects, wherein the antenna may be provided on a vehicle.

A sixth aspect of the present disclosure relates to the antenna design support device of the fifth aspect, wherein the antenna may be provided on a window glass of the vehicle.

A seventh aspect of the present disclosure relates to the antenna design support device of the sixth aspect, wherein the antenna may have a two-dimensional pattern provided on the main surface of the window glass.

An eighth aspect of the present disclosure is a learning device that trains a neural network for predicting the antenna performance of an antenna, and includes an acquisition unit that acquires multiple pieces of graph data representing the shape of the antenna, the antenna performance of which has been determined in advance, and a learning unit that trains a neural network based on the multiple pieces of graph data acquired by the acquisition unit and the antenna performance determined in advance for the antenna, wherein the neural network is a graph neural network for predicting the antenna performance of the antenna based on the graph data.

A ninth aspect of the present disclosure relates to the learning device of the eighth aspect, wherein the graph data may have features including information about the shape of the antenna.

A tenth aspect of the present disclosure is a learning device according to the ninth aspect, wherein the graph data is a graph including nodes representing each point, including the endpoints and intersections, of the antenna lines, and edges representing each line connecting the endpoints or intersections in a connecting relationship, and each of the nodes may have features including position information of each point.

In an eleventh aspect of the present disclosure, in the learning device of the ninth aspect, the graph data is a graph including edges representing each point, including the endpoints and intersections, of the antenna lines, and nodes representing each line connecting the endpoints or intersections in a connecting relationship, and each of the nodes may have a feature including position information of each line.

A twelfth aspect of the present disclosure relates to the learning device of any one of the eighth to eleventh aspects, wherein the antenna may be provided on a vehicle.

A thirteenth aspect of the present disclosure is an antenna design support program for supporting the design of an antenna, which causes a computer to function as an acquisition unit that acquires graph data representing the shape of the antenna, and a prediction unit that uses a pre-trained neural network to predict the antenna performance of the antenna based on the graph data acquired by the acquisition unit, and the neural network is a graph neural network that predicts the antenna performance of the antenna based on the graph data.

A fourteenth aspect of the present disclosure is a learning program for training a neural network to predict the antenna performance of an antenna, the learning program causing a computer to function as an acquisition unit that acquires multiple graph data representing the shape of the antenna, the antenna performance of which has been previously determined, and a learning unit that trains the neural network based on the multiple graph data acquired by the acquisition unit and the antenna performance previously determined for the antenna, wherein the neural network is a graph neural network that predicts the antenna performance of the antenna based on the graph data.

The disclosed technology makes it possible to accurately predict antenna performance.

FIG. 1 is a schematic block diagram of an example of a computer that functions as an antenna design support device and a learning device according to the present embodiment. FIG. 10 is a diagram illustrating an example of the shape of an antenna. FIG. 10 is a diagram showing an example of graph data representing the shape of an antenna. FIG. 1 is a block diagram showing the configuration of a learning device according to an embodiment of the present invention. 1 is a block diagram showing a configuration of an antenna design support apparatus according to an embodiment of the present invention; 4 is a flowchart showing a learning processing routine of the learning device of the present embodiment. 3 is a flowchart showing an antenna design support processing routine of the antenna design support device of the present embodiment. 1A and 1B are diagrams illustrating the shape of an antenna in an embodiment. FIG. 10 is a diagram showing graph data representing the shape of an antenna in an example. FIG. 1 is a diagram illustrating a configuration of a graph neural network in an embodiment.

An example of an embodiment of the disclosed technology will be described below with reference to the drawings. Note that the same or equivalent components and parts in each drawing are given the same reference numerals. Furthermore, the dimensional proportions in the drawings have been exaggerated for the sake of explanation and may differ from the actual proportions. Furthermore, this embodiment will particularly describe in detail the case where an antenna is provided on glass installed in a vehicle, but the present invention is not limited to vehicle antennas.

<Configuration of the learning device according to this embodiment>
FIG. 1 is a block diagram showing the hardware configuration of a learning device 10 according to this embodiment.

As shown in FIG. 1, the learning device 10 has a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, storage 14, an input unit 15, a display unit 16, and a communication interface (I/F) 17. Each component is connected to each other via a bus 19 so that they can communicate with each other.

CPU 11 is a central processing unit that executes various programs and controls each component. That is, CPU 11 reads programs from ROM 12 or storage 14 and executes them using RAM 13 as a work area. CPU 11 controls the above components and performs various arithmetic processing in accordance with the programs stored in ROM 12 or storage 14. In this embodiment, ROM 12 or storage 14 stores a learning program for training the neural network. The learning program may be a single program, or a group of programs consisting of multiple programs or modules.

ROM 12 stores various programs and data. RAM 13 serves as a working area for temporarily storing programs or data. Storage 14 is composed of an HDD (Hard Disk Drive) or SSD (Solid State Drive) and stores various programs, including the operating system, and various data.

The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to enter various inputs.

The input unit 15 accepts as input learning data that is a combination of graph data representing the antenna shape and antenna performance for each of the multiple antennas.

Specifically, graph data is obtained from the shape of an antenna whose performance has been determined in advance. At this time, the graph data has features including information about the antenna shape. More specifically, the graph data is a graph including nodes representing each point, including the endpoints and intersections, of the antenna lines, and edges representing each line connecting the endpoints or intersections in a connecting relationship, and each node has features including position information for each point.

For example, graph data such as that shown in Figure 3 is obtained from the shape of antenna 100 as shown in Figure 2, and the combination of the graph data and the antenna performance previously obtained is used as training data. This training data is prepared for each of the multiple antennas, and the input unit 15 accepts the training data for each of the multiple antennas.

Figure 2 above shows an example of the shape of a conductive antenna 100 mounted on the main surface of a dielectric automobile window glass, i.e., rear windshield. In the example of Figure 2, the power supply section (rectangular closed loop on the left side of Figure 2) and linear antenna element that make up the antenna 100 are indicated by solid lines. In the graph data shown in Figure 3, edges are indicated by solid lines, and nodes representing endpoints and intersections are indicated by numbered circles. Note that the antenna 100 may also be mounted on window glass other than the rear windshield. Furthermore, the antenna 100 is not limited to having a two-dimensional pattern mounted on the main surface of the glass, but may also have a three-dimensional pattern mounted near the glass at a certain distance from the glass. The above "two-dimensional pattern" is intended to include, for example, an antenna pattern mounted along the curved surface of glass on a vehicle.

Measuring equipment is used to measure antenna performance. Indicators of antenna performance include antenna gain, return loss, directional characteristics (half-power angle, front-to-back ratio (F/B)), etc. Specific examples of antenna gain include numerical data on the reception gain (units: [dB], etc.) of radio waves in a specified frequency band. Furthermore, depending on the specifications required of the antenna, the antenna gain may be either the reception gain for vertically polarized waves or the reception gain for horizontally polarized waves, or both.

The display unit 16 is, for example, a liquid crystal display, and displays various information. The display unit 16 may also be a touch panel type, functioning in combination with the input unit 15.

The communication interface 17 is an interface for communicating with other devices, and uses standards such as Ethernet (registered trademark), FDDI (Fiber Distributed Data Interface), and Wi-Fi (registered trademark).

Next, we will explain the functional configuration of the learning device 10. Figure 4 is a block diagram showing an example of the functional configuration of the learning device 10.

As shown in Figure 4, the learning device 10 functionally comprises a learning database (DB) 20, an acquisition unit 22, a learning unit 24, and a model storage unit 26.

The training database 20 stores training data for each of the multiple input antennas.

The acquisition unit 22 acquires multiple pieces of training data from the training database 20.

Based on the received learning data for multiple antennas, the learning unit 24 trains a graph neural network that takes graph data as input and outputs antenna performance.

The model memory unit 26 stores the trained graph neural network.

<Configuration of the antenna design support device according to this embodiment>
FIG. 1 is a block diagram showing the hardware configuration of an antenna design support device 50 according to this embodiment.

As shown in Figure 1 above, the antenna design support device 50 has the same configuration as the learning device 10, and the ROM 12 or storage 14 stores an antenna design support program for predicting antenna performance.

The input unit 15 receives as input graph data generated for the shape of the antenna to be predicted. For example, the input unit 15 receives as input graph data that is a graph including nodes representing the endpoints and intersections of the antenna wires that make up the antenna 100, and edges connecting the endpoints or intersections that are in a connecting relationship, and in which each node has features including position information of the endpoints or intersections.

Next, the functional configuration of the antenna design support device 50 will be described. Figure 5 is a block diagram showing an example of the functional configuration of the antenna design support device 50.

As shown in Figure 5, the antenna design support device 50 functionally comprises a model storage unit 60, an acquisition unit 62, and a prediction unit 64.

The model storage unit 60 stores the graph neural network (GNN) trained by the learning device 10.

The acquisition unit 62 acquires the input graph data. The prediction unit 64 inputs the acquired graph data into a graph neural network and predicts antenna performance.

<Function of the learning device according to this embodiment>
Next, the operation of the learning device 10 according to this embodiment will be described.

Figure 6 is a flowchart showing the flow of the learning process by the learning device 10. The learning process is performed by the CPU 11 in Figure 1 reading a learning program from the ROM 12 or storage 14, expanding it into the RAM 13, and executing it. Furthermore, learning data, which is a combination of graph data representing the antenna shape and antenna performance for each of multiple antennas, is input to the learning device 10 in Figure 4 and stored in the learning database 20.

In step S100, the CPU 11 in FIG. 1 functions as the acquisition unit 22 and acquires multiple pieces of learning data from the learning database 20.

In step S102, the CPU 11 in FIG. 1 functions as the learning unit 24 to learn a graph neural network that inputs graph data and outputs antenna performance based on the received learning data for multiple antennas. The CPU 11 in FIG. 1 stores the learned graph neural network in the model storage unit 26 and terminates the learning process.

<Action of the antenna design support device according to this embodiment>
Next, the operation of the antenna design support device 50 according to this embodiment will be described.

Figure 7 is a flowchart showing the flow of antenna design support processing by the antenna design support device 50. The CPU 11 in Figure 1 reads the antenna design support program from the ROM 12 or storage 14, expands it into the RAM 13, and executes it to perform the antenna design support processing. The model storage unit 60 in Figure 5 stores a trained graph neural network. Graph data generated for the shape of the antenna to be predicted is then input to the antenna design support device 50 in Figure 5.

First, in step S110, the CPU 11 in FIG. 1 functions as the acquisition unit 62 in FIG. 5 to acquire the input graph data.

Next, in step S112, the CPU 11 in FIG. 1 functions as the prediction unit 64 in FIG. 5, inputting the acquired graph data into a graph neural network to predict antenna performance.

Next, in step S114, the CPU 11 in FIG. 1 outputs the prediction results of step S112 above on the display unit 16, and ends the antenna design support process.

<Example>
An example of predicting antenna performance using the learning device 10 and antenna design support device 50 described above will be described.

Figure 8 is a diagram showing the shape of the antenna 100 in Example 1, which is an embodiment of the present invention. Figure 8 shows an example in which the antenna 100 and a defogger 150 including a heating wire for preventing the window from fogging are provided on the main surface of the rear windshield of an automobile. The antenna 100 has two channels (1ch, 2ch).

Here, the directions and origins are defined as follows: When the rear glass is attached to the vehicle body, the long side of the glass plate is the horizontal direction, the short side of the glass plate is the vertical direction, the center of the long side of the glass plate is the horizontal origin, and the center of the horizontal top line of the defogger 150 is the vertical origin.

Figure 9 shows graph data representing the antenna shape in Example 1. Each node represents an end point of an antenna wire or an intersection point between antenna wires.

Functional quantities for each node include the horizontal coordinate, the vertical coordinate, the distance from the left vehicle body to the node, the distance from the right vehicle body to the node, the distance from the upper vehicle body to the node, and the distance from the top line of the defogger to the node.

Here, the vehicle body is basically earthed to metal. Therefore, if an antenna is made by placing a conductive wire on a dielectric, if it gets too close to the vehicle body, capacitive coupling with the vehicle body may occur, causing changes in characteristics. Therefore, in this example, the feature value includes the distance from the vehicle body to the node.

Furthermore, the graphs on the left and right of Figure 9 represent exactly the same information. For example, in both the left and right graphs, node number 0 is connected to node number 1 and node number 42. Similarly, node number 1 is connected to node number 0, node number 2, and node number 5, and all node numbers and their connected node numbers are the same in the left and right graphs. Although the left and right graphs are displayed differently, they represent the exact same node connection relationship information.

Figure 10 is a diagram showing the configuration of the graph neural network in Example 1. Figure 10 shows an example in which the number of input layer neurons is 6, the number of output layer neurons is 4, the number of hidden layers is 6, and the number of hidden layer neurons is 240. The number of input layer neurons corresponds to the number of node features. The number of output layer neurons corresponds to four band average values representing antenna performance (average gain within a specified frequency band of horizontal polarization for 1ch, average gain within a specified frequency band of vertical polarization, average gain within a specified frequency band of horizontal polarization for 2ch, and average gain within a specified frequency band of vertical polarization).

In addition, as a comparative example, Example 2, we use TabNet, a neural network that uses table data as input. The table data has 27 explanatory variables that represent the antenna shape.

Table 1 shows the experimental results for Example 1 (Graph Neural Network GNN) and Example 2 (TabNet). Table 1 also shows the RMSE (Root Mean Squared Error).

1ch_ave_H is the in-band average gain of horizontally polarized waves in the North American FM band (88 MHz to 108 MHz) for the first channel (1ch) supplied from the first power supply, and 1ch_ave_V is the in-band average gain of vertically polarized waves in the North American FM band for the first channel. 2ch_ave_H is the in-band average gain of horizontally polarized waves in the North American FM band for the second channel (2ch) supplied from the second power supply, and 2ch_ave_V is the in-band average gain of vertically polarized waves in the North American FM band for the second channel.

From the above, we can see that using the graph neural network GNN in Example 1 produces better prediction accuracy than Example 2.

As described above, the learning device according to this embodiment acquires multiple graph data representing the shape of an antenna whose antenna performance has been determined in advance, and learns a graph neural network for predicting the antenna performance of the antenna based on the graph data. This makes it possible to learn a graph neural network for accurately predicting antenna performance.

In addition, the antenna design support device according to this embodiment acquires graph data representing the shape of the antenna, and predicts antenna performance based on the graph data using a graph neural network to predict the antenna performance of the antenna. This allows for accurate prediction of antenna performance.

<Modification>
The present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the spirit and scope of the present invention.

For example, while the example has been described in which the learning device and the antenna design support device are configured as separate devices, this is not limited to this, and the learning device and the antenna design support device may also be configured as a single device.

Furthermore, the various processes executed by the CPU after reading software (programs) in each of the above embodiments may be executed by various processors other than the CPU. Examples of processors in this case include dedicated electrical circuits, such as GPUs (Graphics Processing Units), FPGAs (Field-Programmable Gate Arrays), and other PLDs (Programmable Logic Devices) whose circuit configuration can be changed after manufacture, and ASICs (Application Specific Integrated Circuits), which are processors with circuit configurations specifically designed to execute specific processes. Furthermore, the learning process and antenna design support process may be executed by one of these various processors, or by a combination of two or more processors of the same or different types (e.g., multiple FPGAs, or a combination of a CPU and an FPGA, etc.). Furthermore, the hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor devices.

In addition, while the above embodiments have described a configuration in which the learning program and antenna design support program are pre-stored (installed) in storage 14, this is not limiting. The programs may also be provided in a form stored on a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), or a USB (Universal Serial Bus) memory. The programs may also be downloaded from an external device via a network.

Furthermore, while the antenna 100 has been shown mounted on a glass plate such as the rear window of an automobile, this is not a limitation. As long as the antenna 100 is attached to an automotive dielectric, it may be mounted on resin, not just glass. The antenna 100 may be mounted, for example, on an aerodynamic part such as a spoiler, or on a so-called shark fin mounted inside a housing with a protrusion on the roof. Furthermore, when the frequency band of the radio waves to be transmitted and received reaches the high-frequency band of the GHz range, the antenna 100 is not limited to antenna wires arranged two-dimensionally on the main surface of the dielectric, but may have conductors arranged in a solid or mesh pattern in a predetermined planar area. Furthermore, the antenna 100 may have an antenna element with a three-dimensional conductor, and may have specifications compatible with, for example, fifth-generation mobile communication systems (5G), which are compatible with radio waves in the frequency band up to 6 GHz, or radio waves in the quasi-millimeter wave or millimeter wave frequency bands of 6 GHz or above.

Furthermore, while an example has been shown in which graph data is constructed using the endpoints and intersections of antenna lines as nodes and the lines connecting the endpoints and intersections as edges, this is not limitative. Nodes may also be set not only for the endpoints and intersections of antenna lines, but also for intermediate points on the lines. Graph data may also be constructed using the endpoints and intersections of antenna lines as edges and the lines connecting the endpoints and intersections as nodes. In this case, the graph data is a graph that includes edges representing each point, including the endpoints and intersections of the antenna lines, and nodes representing each line connecting the endpoints or intersections in a connecting relationship, and each node has features including position information for each line.

10 Learning device 11 CPU
14 Storage 15 Input unit 16 Display unit 20 Learning database 22 Acquisition unit 24 Learning unit 26 Model storage unit 50 Antenna design support device 60 Model storage unit 62 Acquisition unit 64 Prediction unit 100 Antenna

Claims (14)

An antenna design support device that supports antenna design,
an acquisition unit that acquires graph data representing the shape of the antenna;
a prediction unit that predicts antenna performance of the antenna based on the graph data acquired by the acquisition unit using a pre-trained neural network,
The neural network is a graph neural network for predicting antenna performance of the antenna based on the graph data. The antenna design support device of claim 1, wherein the graph data has features including information about the shape of the antenna. An antenna design support device as described in claim 2, wherein the graph data is a graph including nodes representing each point, including the endpoints and intersections, of the antenna lines, and edges representing each line connecting the endpoints or intersections in a connecting relationship, and each of the nodes has a feature including position information of each point. An antenna design support device as described in claim 2, wherein the graph data is a graph including edges representing each point, including the endpoints and intersections, of the antenna lines, and nodes representing each line connecting the endpoints or intersections in a connected relationship, and each of the nodes has a feature including position information of each line. The antenna design support device according to claim 1, wherein the antenna is installed in a vehicle. The antenna design support device of claim 5, wherein the antenna is mounted on a window glass of the vehicle. The antenna design support device of claim 6, wherein the antenna has a two-dimensional pattern provided on the main surface of the window glass. A learning device that learns a neural network for predicting antenna performance of an antenna, comprising:
an acquisition unit that acquires a plurality of graph data representing the shape of the antenna whose antenna performance has been calculated in advance;
a learning unit that learns a neural network based on the plurality of graph data acquired by the acquisition unit and the antenna performance that is determined in advance for the antenna,
The neural network is a graph neural network for predicting antenna performance of the antenna based on the graph data. The learning device described in claim 8, wherein the graph data has features including information about the shape of the antenna. The learning device described in claim 9, wherein the graph data is a graph including nodes representing each point, including the endpoints and intersections, of the antenna lines, and edges representing each line connecting the endpoints or intersections in a connecting relationship, and each of the nodes has features including position information of each point. The learning device described in claim 9, wherein the graph data is a graph including edges representing each point, including the endpoints and intersections, of the antenna lines, and nodes representing each line connecting the endpoints or intersections in a connecting relationship, and each of the nodes has a feature including position information of each line. The learning device described in claim 8, wherein the antenna is mounted on a vehicle. An antenna design support program for supporting antenna design,
Computer,
an acquisition unit that acquires graph data representing the shape of the antenna; and a prediction unit that predicts antenna performance of the antenna based on the graph data acquired by the acquisition unit using a pre-trained neural network,
The neural network is a graph neural network for predicting antenna performance of the antenna based on the graph data. 1. A learning program for training a neural network to predict antenna performance of an antenna, comprising:
Computer,
an acquisition unit that acquires a plurality of graph data representing the shape of the antenna whose antenna performance has been previously calculated; and a learning unit that learns a neural network based on the plurality of graph data acquired by the acquisition unit and the antenna performance previously calculated for the antenna,
The neural network is a graph neural network for predicting antenna performance of the antenna based on the graph data.