JP7035302B2 - Information processing system, automatic operation device, operation terminal, method and program - Google Patents

Description

The present invention relates to a technique for controlling a ground station that communicates with a controlled object.

The ground station consists of equipment installed on the ground to communicate with the object to be controlled. An artificial satellite is an object to be controlled. Generally, the operator controls the ground station. For example, the operator determines the control content of the ground station based on various determination information, and transmits the control information representing the determined control content to the ground station via the ground station control device. Judgment information includes information indicating the state of each facility of the ground station, environmental information around the ground station, and the like. Further, as each equipment of the ground station, for example, there are transmission / reception equipment, modulation / demodulation equipment, and the like. Further, as the information indicating the state of each equipment, for example, there is an AGC (Automatic Gain Control) gain value obtained from the transmission / reception equipment. Further, as the environmental information around the ground station, for example, there is weather information. Although the operator may make a judgment on the control content based on such judgment information based on a certain standard, it is often based on the experience and intuition of the operator. Further, such a judgment has a problem that there is a possibility of human error and it takes a long time to make a judgment. In addition, if the control content is not appropriate, it takes time and effort to perform control again.

Therefore, there is known a technique for automating the judgment of control contents by an operator by letting a computer perform the judgment instead. For example, the related technique described in Patent Document 1 monitors the ratio of the carrier level and the noise level of the received signal from the artificial satellite, and controls the transmission power of the ground station based on the monitored value.

Further, for example, in the related technology described in Patent Document 2, the inference mechanism unit analyzes the data from each equipment of the ground station and controls each equipment by using the knowledge base unit and the database unit. The knowledge base unit stores information in which the know-how for controlling artificial satellites is systematically organized. The database unit stores information in which various values necessary for controlling artificial satellites are systematically organized.

Further, for example, the related technology described in Patent Document 3 controls the orbit and attitude of each artificial satellite according to the current state of the plurality of artificial satellites. At this time, this related technique optimizes the order of various controls for a plurality of artificial satellites based on the knowledge while learning new knowledge about the control of artificial satellites by using the knowledge base unit and the inference unit. The knowledge base unit is configured to accumulate design data and test data of artificial satellites launched in the past, information on past control of artificial satellites, and the like.

Japanese Unexamined Patent Publication No. 2007-215113 Japanese Unexamined Patent Publication No. 5-181508 Japanese Patent No. 3206907

However, the above-mentioned general technology and related technology have the following problems.

First, there are the following problems in the general technology in which the operator determines the control content of the ground station. It takes time for the operator to make a decision. In addition, if incorrect control is performed, it takes time to correct it. In addition, there are operations that are difficult to operate because it takes too much time for human judgment. For example, if the weather changes drastically and the state of communication quality with the artificial satellite changes in a short time, the operator's judgment cannot be made in time. Also, since the operator is a human being, there is a concern about human error.

Further, in the related technology described in Patent Documents 1 and 2, in order for the computer to judge the control content on behalf of the operator, a control rule of what kind of control content is selected for the judgment information is created in advance. You need to do it. Creating control rules in advance can be difficult for operators.

For example, the related technology described in Patent Document 1 determines whether or not the ratio of the carrier level and the noise level of the received signal is deteriorated from the value in fine weather, whether or not the noise level is higher than the initial value, and the like. Based on this, the rules for how to control the transmission power are set in advance. In this related technology, it can be said that the carrier level and the noise level of the received signal are judgment information. As described above, if the judgment information is small, it is relatively easy to predetermine the control rule. However, as the amount of judgment information increases, the creation of control rules becomes complicated. Therefore, it is difficult to apply this related technology to cases where there is a lot of judgment information.

Further, for the related technology described in Patent Document 2, it is necessary to organize the know-how for controlling artificial satellites in advance and construct a knowledge base unit. Since such know-how is considered to be enormous, it is not easy to organize them in advance.

As described above, in the related technology using the control rule predetermined for automating the operation of the ground station, the work of creating the control rule may be difficult for the operator. In addition, creating a control rule requires meaning, and the reason for the meaning is often ambiguous, such as "experience from past operations." Therefore, it is difficult to optimize the control rule.

For example, when the rainfall is 0 to 100 and is presented to the operator as judgment information, a control rule for transmitting the operation end control when the rainfall is 80 or more is created in advance. At this time, it is necessary to give a meaning to the number "80", which is why it was set to 80 or more. The reason may be clear, but it is often an ambiguous reason such as "experience from past operations". It is difficult to create a control rule based on such an ambiguous reason.

In addition, the related technology described in Patent Document 3 is a plurality of artificial satellites by acquiring new knowledge based on the design data and test data of artificial satellites launched in the past, information on past control, and the like. To control. Here, in order to properly control the artificial satellite, it is necessary to control the ground station so as to optimize the communication state between the ground station and the artificial satellite. For that purpose, it is necessary to consider not only the past data on the artificial satellite and the current state of the artificial satellite, but also the state of the ground station and the environment around the ground station. However, it is difficult for this related technology to accurately control the ground station only based on the knowledge acquired based on the past data on the artificial satellite.

The present invention has been made to solve the above-mentioned problems. That is, an object of the present invention is to provide a technique for automating the control of a device for controlling an object with higher accuracy without the need to set a control rule in advance.

The information processing system of the present invention has a judgment information acquisition unit for acquiring judgment information including information representing the state of a ground station communicating with a controlled object and information representing the environment around the ground station, and the judgment information. A judgment information storage unit that stores the control information, a control information acquisition unit that acquires control information representing the control content for the ground station determined based on the judgment information, a control information storage unit that stores the control information, and the above. Learning data of the ground station control unit that controls the ground station by transmitting control information to the ground station, the judgment information stored in the judgment information storage unit, and the control information stored in the control information storage unit. By applying the learning result stored in the learning result storage unit to the learning unit that performs machine learning, the learning result storage unit that stores the learning result of the machine learning, and the determination information. It includes a control information calculation unit that calculates control information.

Further, in the above-mentioned information processing system, the automatic operation device of the present invention includes the determination information storage unit, the control information storage unit, the learning unit, the learning result storage unit, and the control information calculation unit. Have.

Further, the operation terminal of the present invention has the control information acquisition unit in the above-mentioned information processing system.

Further, the method of the present invention is determined based on the judgment information including the information representing the state of the ground station communicating with the controlled object and the information representing the environment around the ground station, and the judgment information. The control information is calculated by performing machine learning using the control information representing the control content for the ground station as learning data and applying the learning result to the determination information.

Further, the program of the present invention is determined based on the judgment information including the information representing the state of the ground station communicating with the controlled object and the information representing the environment around the ground station, and the judgment information. A machine learning step in which control information representing the control content for the ground station is used as learning data for machine learning, a control information calculation step for calculating the control information by applying the learning result to the judgment information, and a control information calculation step. To let the computer device execute.

INDUSTRIAL APPLICABILITY The present invention can provide a technique for automating the control of a device for controlling an object with higher accuracy without the need to set a control rule in advance.

It is a block diagram which shows the structure of the information processing system as the 1st Embodiment of this invention. It is a block diagram which shows an example of the apparatus which comprises the information processing system as the 1st Embodiment of this invention. It is a figure which shows an example of the hardware composition of each apparatus constituting the information processing system as the 1st Embodiment of this invention. It is a flowchart explaining the operation which the information processing system as the 1st Embodiment of this invention acquires judgment information. It is a flowchart explaining the operation which the information processing system as the 1st Embodiment of this invention acquires the control information and transmits it to a ground station. It is a flowchart explaining the operation which the information processing system as the 1st Embodiment of this invention performs machine learning. It is a flowchart explaining the operation which the information processing system as 1st Embodiment of this invention calculates the control information. It is a block diagram which shows the structure of the information processing system as the 2nd Embodiment of this invention. It is a block diagram which shows an example of the apparatus which comprises the information processing system as the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system acquires the evaluation value in the 1st operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system performs machine learning in the 1st operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system acquires and transmits the control information in the 2nd operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system acquires the evaluation value in the 2nd operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system calculates the control information in the 2nd operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system calculates the control information and sends it to the ground station in the 3rd operation mode of the 2nd Embodiment of this invention. It is a flowchart explaining the operation which the information processing system switches an operation mode in the 2nd Embodiment of this invention. It is a block diagram explaining the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the judgment information stored in the judgment information storage part in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the meteorological information used in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the control information stored in the control information storage part in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the evaluation value stored in the evaluation value storage part in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the structure of the output layer of the learning result constructed in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the structure of another output layer for comparison with the output layer of the learning result constructed in the specific example of the 2nd Embodiment of this invention. It is a figure explaining an example of the structure of the other output layer for comparison with the output layer of the learning result constructed in the specific example of the 2nd Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(First Embodiment)
FIG. 1 shows a functional block configuration of the information processing system 1 as the first embodiment of the present invention. In FIG. 1, the information processing system 1 includes a judgment information acquisition unit 101, a judgment information storage unit 102, a control information acquisition unit 103, a control information storage unit 104, a ground station control unit 105, and a learning unit 106. It includes a learning result storage unit 107 and a control information calculation unit 108. The information processing system 1 is a system that controls a ground station that communicates with a controlled object. The ground station is composed of various facilities.

Here, an example of the hardware configuration of the information processing system 1 will be described with reference to FIGS. 2 and 3. As shown in FIG. 2, each functional block of the information processing system 1 can be distributed and configured to be a ground station control device 300, an automatic operation device 400, and an operation terminal 500. Each of these devices is communicably connected to each other via a network. For example, the ground station control device 300 includes a determination information acquisition unit 101 and a ground station control unit 105. Further, the automatic operation device 400 includes a judgment information storage unit 102, a control information storage unit 104, a learning result storage unit 107, a learning unit 106, and a control information calculation unit 108. Further, the operation terminal 500 includes a control information acquisition unit 103.

In this case, for example, as shown in FIG. 3, the ground station control device 300 includes a CPU (Central Processing Unit) 3001, a memory 3002, and a computer device including a network interface 3005. The memory 3002 is composed of a RAM (Random Access Memory), a ROM (Read Only Memory), an auxiliary storage device (hard disk, etc.) and the like. The network interface 3005 is an interface for connecting to a network. Further, the automatic operation device 400 is composed of a computer device including a CPU 4001, a memory 4002, and a network interface 4005. The memory 4002 is composed of a RAM, a ROM, an auxiliary storage device, and the like. The network interface 4005 is an interface for connecting to a network. Further, the operation terminal 500 is composed of a computer device including a CPU 5001, a memory 5002, an output device 5003, an input device 5004, and a network interface 5005. The memory 5002 is composed of a RAM, a ROM, an auxiliary storage device, and the like. The output device 5003 is composed of a device that outputs information, such as a display device and a printer. The input device 5004 is configured by a device that receives input of a user operation, such as a keyboard and a mouse. The network interface 5005 is an interface for connecting to a network.

In this case, the determination information acquisition unit 101 and the ground station control unit 105 are composed of the network interface 3005 of the ground station control device 300 and the CPU 3001 that reads and executes the program and various data stored in the memory 3002. Further, the determination information storage unit 102, the control information storage unit 104, and the learning result storage unit 107 are configured by the memory 4002 of the automatic operation device 400. Further, the learning unit 106 is composed of a CPU 4001 that reads and executes a program and various data stored in the memory 4002 of the automatic operation device 400. Further, the control information calculation unit 108 is composed of a network interface 4005 of the automatic operation device 400 and a CPU 4001 that reads and executes a program and various data stored in the memory 4002. Further, the control information acquisition unit 103 is composed of a network interface 5005 of the operation terminal 500, an output device 5003, an input device 5004, and a CPU 5001 that reads and executes a program and various data stored in the memory 5002. However, the hardware configuration of the information processing system 1 and its functional blocks is not limited to the above configuration.

Next, the details of each functional block will be described.

The judgment information acquisition unit 101 acquires the judgment information. Judgment information includes information representing the state of the ground station and information representing the environment around the ground station. For example, the determination information acquisition unit 101 acquires information indicating the state from each facility constituting the ground station. Further, for example, the determination information acquisition unit 101 acquires information representing the environment from ancillary equipment that detects the environment around the ground station.

The judgment information storage unit 102 stores the judgment information acquired by the judgment information acquisition unit 101.

The control information acquisition unit 103 acquires control information determined based on the determination information. The control information represents the control content for the ground station. For example, the control information acquisition unit 103 may output the determination information to the output device 5003 and acquire the information input from the input device 5004 according to the output as the control information. Such control information typically represents the control content determined by the operator's judgment based on the presented judgment information.

The control information storage unit 104 stores the control information acquired by the control information acquisition unit 103.

The control information of the control information storage unit 104 and the judgment information of the judgment information storage unit 102 are associated with each other so that which control information is determined based on which judgment information can be collated. For example, in the control information storage unit 104, it is assumed that identification information is added to the control information and stored. In this case, the determination information storage unit 102 stores the determination information in association with the identification information of the control information determined using the determination information.

Alternatively, the determination information storage unit 102 may add identification information to the determination information and store it. In this case, the control information storage unit 104 stores the control information in association with the identification information of the determination information used for the determination.

The judgment information and the control information are not limited to these, and may be stored together with any information in any format as long as it is possible to collate which control information is determined based on which judgment information. ..

The ground station control unit 105 controls the ground station by transmitting the control information acquired by the control information acquisition unit 103 to the ground station.

The learning unit 106 performs machine learning using the judgment information stored in the judgment information storage unit 102 and the control information stored in the control information storage unit 104 as learning data. Specifically, the learning unit 106 may generate an operation as a learning result in which each stored determination information is input and control information determined based on the determination information is output. Then, the learning unit 106 stores the learning result of machine learning in the learning result storage unit 107.

The control information calculation unit 108 calculates the control information by applying the learning result stored in the learning result storage unit 107 to the determination information. Here, the control information calculation unit 108 reads one or a plurality of recently stored judgment information from the judgment information storage unit 102 as the judgment information to which the learning result is applied. Such one or more judgment information may be acquired within a predetermined period up to the time when the calculation is performed, for example.

Further, the control information calculation unit 108 outputs the calculated control information. The output destination may be the ground station control unit 105. In this case, the ground station control unit 105 transmits the control information calculated by the control information calculation unit 108 to the ground station.

The operation of the information processing system 1 configured as described above will be described with reference to the drawings.

First, FIG. 4 shows an operation in which the information processing system 1 acquires determination information.

In FIG. 4, first, the judgment information acquisition unit 101 acquires the judgment information (step A11). For example, the judgment information is transmitted from each facility constituting the ground station. In addition, the judgment information is transmitted from ancillary equipment that detects the environment around the ground station.

Next, the judgment information acquisition unit 101 stores the acquired judgment information in the judgment information storage unit 102 (step A12). The determination information stored in this step is not associated with the control information at this point.

The information processing system 1 repeats the above operation.

Next, FIG. 5 shows an operation in which the information processing system 1 acquires control information and transmits it to a ground station.

In FIG. 5, first, the control information acquisition unit 103 reads the determination information used for determining the control information from the determination information storage unit 102 (step B11).

For example, it is assumed that the control content is determined based on one or more judgment information received in the last N seconds. N is a positive number. In this case, the control information acquisition unit 103 may read the determination information added in the last N seconds in the determination information storage unit 102.

Next, the control information acquisition unit 103 acquires the control information determined based on the determination information acquired in step B11 (step B12).

As described above, the control information acquisition unit 103 may output the determination information corresponding to the output device 5003 and acquire the control information input via the input device 5004 according to the output.

Next, the control information acquisition unit 103 outputs the acquired control information to the ground station control unit 105. Then, the ground station control unit 105 controls the ground station by transmitting control information to the ground station (step B13).

Next, the control information acquisition unit 103 stores the acquired control information in the control information storage unit 104 (step B14).

Next, the control information acquisition unit 103 associates the determination information read in step B11 with the control information acquired in step B12 (step B15). As described above, for example, when the identification information is added to the control information, the identification information of the control information acquired in step B12 may be added and associated with the determination information acquired in step B11.

The information processing system 1 repeats the above operation.

Next, FIG. 6 shows an operation in which the information processing system 1 performs learning.

In FIG. 6, first, the learning unit 106 collates the information stored in the judgment information storage unit 102 and the control information storage unit 104 to generate learning data (step C11).

Here, the learning data is a set of judgment information and control information determined based on the judgment information. The learning data does not necessarily have to be a set consisting of judgment information and control information one by one. For example, it is assumed that the control content is determined based on one or more judgment information received in the last N seconds. Further, it is assumed that it is stipulated that a plurality of types of control contents are determined at one time. In this case, the learning data may be a set consisting of one or more determination information and one or more control information.

Next, the learning unit 106 performs machine learning based on the learning data (step C12).

Specifically, the learning unit 106 derives an operation in which the judgment information of the learning data is used as an input signal, the control information that is a set of the judgment information is used as a teacher signal, and the teacher signal is output based on the input signal. Just do it.

Next, the learning unit 106 stores the learning result of machine learning in the learning result storage unit 107 (step C13).

With the above, the information processing system 1 ends the operation of performing learning.

Next, FIG. 7 shows an operation in which the information processing system 1 calculates control information.

In FIG. 7, first, the control information calculation unit 108 reads the judgment information used for determining the control information from the judgment information storage unit 102 (step D11).

For example, when the control content is determined based on one or more judgment information received in the last N seconds, the control information calculation unit 108 is added in the judgment information storage unit 102 in the last N seconds. All you have to do is read the judgment information.

Next, the control information calculation unit 108 calculates and outputs control information by applying the learning result stored in the learning result storage unit 107 to the read determination information (step D12).

As described above, when the output destination is the ground station control unit 105, the ground station control unit 105 transmits the control information calculated by the control information calculation unit 108 to the ground station.

With the above, the information processing system 1 ends the operation of calculating the control control information.

Next, the effect of the first embodiment of the present invention will be described.

The information processing system as the first embodiment of the present invention can automate the control of the device for controlling the object with higher accuracy without the need to set the control rule in advance.

The reason will be explained. In the present embodiment, the judgment information acquisition unit acquires judgment information including information representing the state of the ground station communicating with the controlled object and information representing the environment around the ground station. The acquired judgment information is stored in the judgment information storage unit. In addition, the control information acquisition unit acquires control information representing the control content for the ground station determined based on the determination information. The acquired control information is stored in the control information storage unit. Then, the ground station control unit transmits the acquired control information to the ground station to control the ground station. Then, the learning unit performs machine learning using the judgment information stored in the judgment information storage unit and the control information stored in the control information storage unit as learning data. The learning result of machine learning is stored in the learning result storage unit. This is because the control information calculation unit calculates the control information by applying the learning result stored in the learning result storage unit to the determination information.

As described above, in this embodiment, machine learning is performed using a set of judgment information representing the state of the ground station and the environment around the ground station and control information determined by the operator based on the judgment information as learning data. , There is no need to create control rules in advance. And, in this embodiment, since the control information is calculated based on the judgment information by using the learning result of such machine learning, the control of the ground station similar to that of the veteran operator can be performed more than the veteran operator. It can be judged and executed in a short time. Further, in the present embodiment, since the control information is calculated based on the judgment information by using the learning result of such machine learning, the human error that may occur in the control based on the judgment of the operator does not occur.

That is, in this embodiment, learning results, which are control rules that are difficult to create manually, can be automatically created by machine learning by the learning unit without requiring meaning or the like. Further, in the present embodiment, the judgment information acquisition unit and the control information acquisition unit acquire and store a sufficient amount of learning data to generate the learning result of the machine learning which is such a control rule. As described above, this embodiment does not require human cost for creating control rules and collecting learning data.

Further, in this embodiment, machine learning is performed based not only on past data and past control information on the object to be controlled, but also on various judgment information representing the state of the ground station and the environment around the ground station. Therefore, control rules for calculating control information with high accuracy are automatically created according to changes in the situation surrounding the ground station. Therefore, in the present embodiment, the control information calculation unit calculates the control information using the control rule generated in this way, so that the operator can quickly determine the control content, which is difficult for the operator. For example, even when the state of the ground station or the environment around the ground station changes suddenly, the present embodiment can quickly determine the control content and transmit the control information to the ground station in response to the sudden change. .. As a result, the present embodiment optimizes the state of the ground station and improves the communication quality with the controlled object with higher accuracy.

Although the example in which the output destination of the control information by the control information calculation unit is the ground station control unit has been described in step D12 of FIG. 7, the output destination may be the control information acquisition unit. When the output destination is the control information acquisition unit, the control information acquisition unit outputs the calculated control information as reference information to the output device together with the determination information used for determining the control information in step B12 of FIG. Then, the control information acquisition unit may acquire the information input from the input device in response to the output of these information as control information. In this case, such control information represents the control content determined by the operator based on the determination information and the control information presented as reference information. As described above, this embodiment is suitable not only for the purpose of automating the operation by the control information calculation unit, but also for the use of presenting the calculation result by the control information calculation unit to the operator to support the operator's judgment. be.

(Second embodiment)
Next, a second embodiment of the present invention will be described in detail with reference to the drawings. In each of the drawings referred to in the description of the present embodiment, the same components as those of the first embodiment of the present invention and the steps operating in the same manner are designated by the same reference numerals to provide detailed details of the present embodiment. The explanation is omitted.

First, FIG. 8 shows the configuration of the information processing system 2 as the second embodiment of the present invention. The information processing system 2 has a control information acquisition unit 203 instead of the control information acquisition unit 103 and a learning unit 206 instead of the learning unit 106 for the information processing system 1 as the first embodiment of the present invention. The difference is that the control information calculation unit 208 is provided in place of the control information calculation unit 108. Further, it is different in that the evaluation value acquisition unit 209, the evaluation value storage unit 210, and the operation mode switching unit 211 are further provided.

Here, an example of the hardware configuration of the information processing system 2 will be described. As shown in FIG. 9, each functional block of the information processing system 2 can be distributed and configured to be a ground station control device 300, an automatic operation device 410, and an operation terminal 510. Each of these devices is communicably connected to each other via a network. The configuration of the ground station control device 300 is the same as that of the first embodiment of the present invention described with reference to FIG. Further, the automatic operation device 410 includes a judgment information storage unit 102, a control information storage unit 104, a learning result storage unit 107, an evaluation value storage unit 210, a learning unit 206, and a control information calculation unit 208. Further, the operation terminal 510 includes a control information acquisition unit 203, an evaluation value acquisition unit 209, and an operation mode switching unit 211.

In this case, the ground station control device 300, the automatic operation device 410, the operation terminal 510, and each functional block arranged in each device are the same as the first embodiment of the present invention described with reference to FIG. It consists of similar hardware elements. Further, the evaluation value storage unit 210 is configured by the memory 4002 of the automatic operation device 410. Further, the evaluation value acquisition unit 209 and the operation mode switching unit 211 are configured as follows. That is, each of these functional blocks is composed of the network interface 5005 of the operation terminal 510, the output device 5003, the input device 5004, and the CPU 5001 that reads and executes the program and various data stored in the memory 5002. However, the hardware configuration of the information processing system 2 and its functional blocks is not limited to the above configuration.

Next, the details of each functional block will be described.

In the first operation mode described later, the control information acquisition unit 203 is configured in the same manner as the control information acquisition unit 103 in the first embodiment of the present invention.

Further, the control information acquisition unit 203 is configured in substantially the same manner as the control information acquisition unit 103 in the first embodiment of the present invention in the second operation mode described later, except for the following points. That is, in the second operation mode, the control information acquisition unit 203 is based on the judgment information read from the judgment information storage unit 102 and the control information calculated by the control information calculation unit 208 based on the read judgment information. Acquire the determined control information. For example, the control information acquisition unit 203 may output the determination information read from the determination information storage unit 102 and the control information calculated by the control information calculation unit 208 based on the read determination information to the output device 5003. .. Then, the control information acquisition unit 203 may acquire the information input from the input device 5004 according to the output as the control information.

Further, in the second operation mode, the control information acquisition unit 203 stores the control information acquired by the input in the control information storage unit 104, and also stores the control information calculated by the control information calculation unit 208 as control information. It is stored in the storage unit 104. Then, in addition to associating the judgment information read from the judgment information storage unit 102 with the control information acquired by input, the control information acquisition unit 203 has the judgment information read from the judgment information storage unit 102 and the control information calculation unit. It is associated with the control information calculated by 208.

Further, the control information acquisition unit 203 is configured not to function in the third operation mode described later.

The evaluation value acquisition unit 209 acquires the evaluation value for the control information and stores it in the evaluation value storage unit 210. For example, the evaluation value acquisition unit 209 may acquire such an evaluation value from the input device 5004 of the operation terminal 510. Further, the evaluation value acquisition unit 209 stores the corresponding evaluation value in the evaluation value storage unit 210 in association with the identification information of the control information to be evaluated.

The control information to be evaluated is slightly different between the first operation mode and the second operation mode. In the first operation mode, the control information to be evaluated is the control information acquired by the control information acquisition unit 203 and transmitted to the ground station. It can be said that the evaluation value for such control information represents the evaluation for the control result based on the judgment by the operator. In the second operation mode, in addition to the control information acquired by the control information acquisition unit 203 and transmitted to the ground station, the control information calculated by the control information calculation unit 208 is the evaluation target. It can be said that the evaluation value for such control information represents the evaluation regarding the accuracy of the learning result.

Further, the evaluation value acquisition unit 209 is configured not to function in the third operation mode.

The learning unit 206 is configured in substantially the same manner as the learning unit 106 in the first embodiment of the present invention, but has a point of performing machine learning for each type of control and a point of narrowing down information to be adopted as learning data. different.

Specifically, the learning unit 206 sets learning data as a set of control information of the type and one or more judgment information used when determining the control information for each type of control represented by the control information. Generate as. Then, the learning unit 206 performs machine learning using the learning data including the control information that is the evaluation target of the evaluation value satisfying the predetermined condition among such learning data. For example, it is assumed that the evaluation value is defined as a numerical value indicating that the larger the value, the higher the evaluation. In this case, the predetermined condition may be that the evaluation value is equal to or higher than the threshold value. Then, the learning unit 206 stores the learning result for each type of control in the learning result storage unit 107.

Further, the learning unit 206 is configured to function in the first operation mode and the second operation mode, and not to function in the third operation mode.

The control information calculation unit 208 calculates and outputs control information by applying the learning result for each type of control to the judgment information. The output destination differs between the second operation mode and the third operation mode.

In the second operation mode, the control information calculation unit 208 outputs the calculation result of the control information to the control information acquisition unit 203. Further, in the third operation mode, the control information calculation unit 208 outputs the calculation result of the control information to the ground station control unit 105.

Further, the control information calculation unit 208 is configured to function in the second operation mode and the third operation mode, and not to function in the first operation mode.

The operation mode switching unit 211 executes switching to any one of the first operation mode, the second operation mode, and the third operation mode.

The first operation mode is an operation mode in which the control information acquisition unit 203 and the learning unit 206 function and the control information calculation unit 208 does not function. In the first operation mode, the control information acquisition unit 203 acquires control information input by an operator or the like. For example, in the first operation mode, the control information acquisition unit 203 outputs the read determination information to the output device 5003, acquires the control information input from the input device 5004 in response to the output, and obtains the ground station control unit. It may be sent to 105. Further, in the first operation mode, the evaluation value acquisition unit 209 acquires the evaluation value for the control information input by the operator. Then, the learning unit 206 performs machine learning using the combination of the determination information and the control information input by the operator as the learning data whose evaluation value satisfies a predetermined condition. That is, the first operation mode is suitable as a mode in which machine learning for automation is performed while being operated by the operator at the initial stage of introducing the automatic operation device 410.

The second operation mode is an operation mode in which the control information acquisition unit 203, the learning unit 206, and the control information calculation unit 208 function together. In the second operation mode, the control information acquisition unit 203 acquires both the control information input by the operator and the like and the control information from the control information calculation unit 208. For example, the control information acquisition unit 203 outputs the read determination information and the control information calculated by the control information calculation unit 208 based on the determination information to the output device 5003. Then, the control information acquisition unit 203 may acquire the control information input from the input device 5004 in response to the output of these information and transmit it to the ground station control unit 105. Further, the control information calculation unit 208 calculates the control information in response to the request of the control information acquisition unit 203, and outputs the calculation result to the control information acquisition unit 203. Further, the evaluation value acquisition unit 209 acquires and stores the evaluation value for the control information input by the operator and the evaluation value for the control information calculated by the control information calculation unit 208. Then, the learning unit 206 performs machine learning using a combination of the judgment information and the control information input or calculated by the operator whose evaluation value satisfies a predetermined condition as learning data. .. That is, the second operation mode is suitable as a mode for confirming the accuracy of the machine learning result and further improving the accuracy before applying the machine learning result in the first operation mode to the formal operation.

Further, the third operation mode is an operation mode in which the control information acquisition unit 203 and the learning unit 206 do not function, but the control information calculation unit 208 functions. In the third operation mode, the control information calculation unit 208 transmits the control information calculated based on the read determination information to the ground station control unit 105. The transmitted control information is transmitted to the ground station by the ground station control unit 105. That is, the third operation mode is suitable as a mode in which the control of the ground station is automated by using the learning results of machine learning in the first and second operation modes.

For example, the operation mode switching unit 211 may switch the operation mode based on the information input via the input device 5004 of the operation terminal 510. Alternatively, the operation mode switching unit 211 may switch the operation mode according to a predetermined schedule.

The operation of the information processing system 2 configured as described above will be described with reference to the drawings. In the following description of the operation, when information is exchanged between functional blocks arranged in different devices, it is assumed that the devices are communicating with each other, and an explicit description thereof will be omitted.

First, the operation of the information processing system 2 in the first operation mode will be described.

In the first operation mode, the operation of the information processing system 2 to acquire the determination information is the same as that of the first embodiment of the present invention described with reference to FIG.

Further, in the first operation mode, the operation of the information processing system 2 to acquire the control information and transmit it to the ground station is the same as the first embodiment of the present invention described with reference to FIG.

Next, FIG. 10 shows an operation in which the information processing system 2 acquires an evaluation value in the first operation mode.

In FIG. 10, first, the evaluation value acquisition unit 209 outputs the control information acquired in step B12 of FIG. 5 (step E21). At this time, the evaluation value acquisition unit 209 may output the control information together with the information representing the result of the control of the ground station by the control information. The output destination may be, for example, the output device 5003.

Next, the evaluation value acquisition unit 209 acquires the evaluation value for the control information (step E22). For example, the evaluation value acquisition unit 209 may acquire the evaluation value via the input device 5004.

Next, the evaluation value acquisition unit 209 stores the acquired evaluation value in the evaluation value storage unit 210 in association with the identification information of the control information to be evaluated (step E23).

With the above, the information processing system 2 ends the operation of acquiring the evaluation value.

The evaluation value acquisition operation shown in FIG. 10 may be executed following the control information acquisition operation shown in FIG. As a result, the operator can input the control information and the evaluation value in succession, and the burden on the operator is reduced.

Next, FIG. 11 shows an operation in which the information processing system 2 learns in the first operation mode.

In FIG. 11, first, the learning unit 206 collates the information stored in the judgment information storage unit 102 and the control information storage unit 104, respectively, and determines the control information and one or more judgments used for determining the control information. A set with information is generated as learning data (step C21).

Next, the learning unit 206 extracts the learning data including the evaluation information associated with the evaluation value satisfying the predetermined condition from the learning data generated in step C21 by referring to the evaluation value storage unit 210 (step C22). ).

Next, the learning unit 206 executes the operations of steps C23 to C24 for each type of control.

Here, first, the learning unit 206 uses the extracted learning data to perform machine learning using judgment information as an input and control information of the corresponding type as a teacher signal (step C23).

Next, the learning unit 206 stores the learning result of machine learning in the learning result storage unit 107 (step C24).

When the above operation is completed for all types of control information, the information processing system 2 ends the learning operation.

This is the end of the description of the operation of the information processing system 2 in the first operation mode.

Next, the operation of the information processing system 2 in the second operation mode will be described.

In the second operation mode, the operation of the information processing system 2 to acquire the determination information is the same as that of the first embodiment of the present invention described with reference to FIG.

Next, FIG. 12 shows an operation in which the information processing system 2 acquires control information and transmits it to the ground station in the second operation mode.

In FIG. 12, the control information acquisition unit 203 executes step B11 in the same manner as in the first embodiment of the present invention, and acquires the determination information used for determining the control information from the determination information storage unit 102.

Next, the control information acquisition unit 203 outputs the determination information acquired in step B11 to the control information calculation unit 208, and as a result, acquires the control information calculated by the control information calculation unit 208 (step B21).

Next, the control information acquisition unit 203 acquires the control information determined based on the read determination information and the control information calculated by the control information calculation unit 208 (step B22). For example, the control information acquisition unit 203 may output the read determination information and the calculated control information to the output device 5003, and may acquire the information input from the input device 5004 in response to the output as control information. ..

Next, the control information acquisition unit 203 executes step B13 in the same manner as in the first embodiment of the present invention. As a result, the acquired control information is transmitted to the ground station.

Next, the control information acquisition unit 203 stores the control information calculated by the control information calculation unit 208 in step B21 and the control information acquired in step B22 in the control information storage unit 104 (step B24).

Next, the control information acquisition unit 203 executes step B15 in the same manner as in the first embodiment of the present invention. As a result, the control information is associated with the judgment information used for the determination or calculation.

With the above, the information processing system 2 ends the operation of acquiring the control information and transmitting it to the ground station.

Next, FIG. 13 shows an operation in which the information processing system 2 acquires an evaluation value in the second operation mode.

In FIG. 13, first, the evaluation value acquisition unit 209 executes steps E21 to E23 shown in FIG. As a result, the evaluation value acquisition unit 209 acquires and stores the evaluation value for the control information acquired by the input in step B22 of FIG.

Next, the evaluation value acquisition unit 209 outputs the control information calculated in step B21 shown in FIG. 12 (step E24).

Next, the evaluation value acquisition unit 209 acquires the evaluation value for the calculated control information (step E25). For example, the evaluation value acquisition unit 209 may acquire the evaluation value via the input device 5004.

Next, the evaluation value acquisition unit 209 stores the acquired evaluation value in the evaluation value storage unit 210 in association with the calculated identification information of the control information (step E26).

With the above, the information processing system 2 ends the operation of acquiring the evaluation value.

The operation in which the information processing system 2 learns in the second operation mode using the judgment information, control information, and evaluation value stored in this way is described in the first operation mode described with reference to FIG. It is the same as the learning operation.

Next, FIG. 14 shows an operation in which the information processing system 2 calculates control information in the second operation mode. This operation is started by being requested by the control information acquisition unit 203 in step B21 shown in FIG.

In FIG. 14, first, the control information calculation unit 208 acquires the determination information used for calculating the control information from the control information acquisition unit 203 (step D21).

Next, the control information calculation unit 208 applies the learning result for each type of control to the acquired determination information, and calculates the control information (step D22). Then, the control information calculation unit 208 outputs the calculated control information to the control information acquisition unit 203.

This completes the description of the operation in which the information processing system 2 calculates the control information.

Next, the operation of the information processing system 2 in the third operation mode will be described.

In the third operation mode, the operation of the information processing system 2 to acquire the determination information is the same as that of the first embodiment of the present invention described with reference to FIG.

Further, in the third operation mode, the control information acquisition unit 203 and the learning unit 206 do not operate.

Here, in the third operation mode, the operation in which the information processing system 2 calculates the control information and transmits it to the ground station will be described with reference to FIG.

Here, first, the control information calculation unit 208 reads the judgment information used for calculating the control information from the judgment information storage unit 102 (step D31).

Next, the control information calculation unit 208 applies the learning result for each type of control to the read determination information, and calculates the control information. Then, the control information calculation unit 208 transmits the calculated control information to the ground station control unit 105 (step D32).

Here, if the content of the control information does not change the current status of the ground station (No in step D33), the information processing system 2 ends the process related to this type of control information.

On the other hand, when the content of the control information is a content that changes the current status of the ground station (No in step D33), the ground station control unit 105 transmits the control information to the ground station (step D34).

When the processing of steps D32 to D34 for all types of control information is completed, the information processing system 2 determines whether or not to terminate the operation (step D35).

If the operation is not terminated, the information processing system 2 repeats the process from step D31. When the operation is terminated, the information processing system 2 terminates the operation.

This completes the description of the operation in which the information processing system 2 calculates the control information and transmits it to the ground station.

This completes the description of the operation of the information processing system 2 in the first to third operation modes.

Next, FIG. 16 shows an operation in which the information processing system 2 switches the operation mode.

In FIG. 16, first, the operation mode switching unit 211 of the operation terminal 410 acquires information for designating the operation mode (step F21). The information that specifies the operation mode is information that represents any of the first, second, and third operation modes. Information specifying the operation mode may be acquired via the input device 5004.

Here, the case where the information for designating the first operation mode is acquired (the first operation mode in step F22) will be described. In this case, the operation mode switching unit 211 controls the control information calculation unit 208, the learning unit 206, the judgment information acquisition unit 101, and the evaluation value acquisition unit 209 as follows (step F23).

That is, in this case, the operation mode switching unit 211 controls the control information acquisition unit 203 to execute the control information acquisition operation shown in FIG. Further, the operation mode switching unit 211 controls the evaluation value acquisition unit 209 to execute the evaluation value acquisition operation shown in FIG. Further, the operation mode switching unit 211 controls the learning unit 206 to execute the learning operation shown in FIG. Further, the operation mode switching unit 211 controls the control information calculation unit 208 so as not to function.

As a result, the information processing system 2 is in a mode in which the learning unit 206 learns the judgment of the operator while performing the operation using the control information determined by the judgment of the operator based on the judgment information.

Further, a case where information for designating the second operation mode is acquired (second operation mode in step F22) will be described. In this case, the operation mode switching unit 211 controls the control information calculation unit 208, the learning unit 206, the judgment information acquisition unit 101, and the evaluation value acquisition unit 209 as follows (step F24).

That is, in this case, the operation mode switching unit 211 controls the control information acquisition unit 203 to execute the control information acquisition operation shown in FIG. 12. Further, the operation mode switching unit 211 controls the evaluation value acquisition unit 209 to execute the evaluation value acquisition operation shown in FIG. Further, the operation mode switching unit 211 controls the learning unit 206 to execute the learning operation shown in FIG. Further, the operation mode switching unit 211 controls the control information calculation unit 208 to execute the control information calculation operation shown in FIG.

As a result, the information processing system 2 presents the control information calculated based on the learning result to the operator as a proposal, and finally operates using the control information determined by the operator, while learning unit 206. It becomes a mode to continue learning by.

Further, the case where the information for designating the third operation mode is acquired (the third operation mode in step F22) will be described. In this case, the operation mode switching unit 211 controls the control information calculation unit 208, the learning unit 206, the judgment information acquisition unit 101, and the evaluation value acquisition unit 209 as follows (step F25).

That is, in this case, the operation mode switching unit 211 controls the control information acquisition unit 203 and the learning unit 206 so as not to function. Further, the operation mode switching unit 211 controls the evaluation value acquisition unit 209 so as not to function. Further, the operation mode switching unit 211 controls the control information calculation unit 208 to execute the control information calculation operation shown in FIG.

As a result, the information processing system 2 becomes a mode in which the operation is automated by using the control information calculated by using the learning results adjusted by the first and second operation modes.

Next, the operation of the information processing system 2 will be shown by a specific example.

In this specific example, it is assumed that the object to be controlled is an artificial satellite. Further, as shown in FIG. 17, the information processing system 2 is connected to the transmission / reception equipment and the modulation / demodulation equipment of the ground station provided on the ground corresponding to the artificial satellite. Further, the information processing system 2 is connected to ancillary equipment that detects the weather around the ground station.

Here, the judgment information includes the AGC (Automatic Gain Control) reception level and the current loop band transmitted from the transmission / reception equipment, the current bit rate transmitted from the modulation / demodulation equipment, and the current weather information transmitted from the incidental equipment. It is said that there is weather forecast information. These determination information is transmitted to the ground station control device 300.

Further, it is assumed that there are three types of control information: loop band setting control for transmission / reception equipment, bit rate setting control for modulation / demodulation equipment, and control for terminating operation.

First, FIG. 18 shows an example of judgment information stored in the judgment information storage unit 102 by the operation shown in FIG. In FIG. 18, each row represents an entry of judgment information acquired at the same time. Each entry is composed of a tag, an acquisition date and time, and various judgment information described above.

In each entry, the acquisition date and time represents the date and time when the judgment information of the entry was acquired. Further, as for the AGC reception level, the current loop band, and the current bit rate, the acquired values are stored as they are. Further, the current weather information and the weather forecast information are stored with values based on the correspondence table shown in FIG. The current weather information represents the weather at the time the information was acquired. In addition, the weather forecast information represents a weather forecast at predetermined intervals from the time when the information is acquired to a predetermined period. In the example of FIG. 19, in each entry, the weather forecast information consists of a total of four numerical values representing the weather forecast every 30 minutes from the time when the information is acquired to 2 hours later.

Further, in each entry, the tag represents the identification information of the control information determined by using the judgment information of the entry. Here, the identification information of the control information is the control date and time. That is, the tag stores the control date and time of the corresponding entry in the control information storage unit 104, which will be described later. The tag may not be stored yet at the time when the entry of the judgment information is created. The tag is stored, for example, in step B15 shown in FIG. 5 or FIG.

In this example, it is assumed that the judgment information acquired in the latest N seconds at the time when the process for determining the control information is executed is used for determining the control information. For example, in FIG. 18, the control date and time “xxx-xx xx: xx”, which is the identification information of the same control information, is stored as a tag in each entry in the first to fifth lines. That is, the judgment information in the first to fifth lines is the judgment information used when the control information having the control date and time "xxx-xx xx: xx" as the identification information is determined.

Next, FIG. 20 shows an example of the control information stored in the control information storage unit 104 by the operation shown in FIG. 5 or FIG. In FIG. 20, each line represents an entry of control information transmitted by the control information acquisition unit 203 to the ground station via the ground station control device 300. Each entry consists of a control date and time, a bit rate set value, a loop band set value, and an operation end control value. The control date and time represents the date and time when this control information was acquired or calculated. Further, the bit rate set value represents a control for setting the bit rate of the modulation / demodulation equipment to the value. Further, the loop band set value represents control for setting the loop band of the transmission / reception equipment to the value. Further, as for the operation end control value, 1 indicates off (operation end) and 0 indicates on (operation continuation).

Next, FIG. 21 shows an example of the evaluation value stored in the evaluation value storage unit 210 by the operation shown in FIG. 10 or 13. In FIG. 21, each row represents an entry for the evaluation value. Each entry consists of a tag and an evaluation value. The tag represents the identification information (here, the control date and time) of the control information to be evaluated. The evaluation value is defined here to take any integer from 0 to 100. 100 represents the best rating and 0 represents the worst rating. In the evaluation value, the value input by the operator or the like in step E22 or E25 is stored.

Next, in such a specific example, the operation of the information processing system 2 for learning will be described. This operation is executed in the first and second operation modes.

First, the learning unit 206 collates the judgment information storage unit 102 of FIG. 18 with the control information storage unit 104 of FIG. 20, and obtains control information and one or more judgment information in which the control date and time are stored as tags. Is generated as training data (step C21).

Next, the learning unit 206 extracts an entry whose evaluation value is lower than the threshold value (for example, 50) by referring to the control information storage unit 104 of FIG. 21, and steps the learning data including the control information indicated by the tag. It is deleted from the training data generated in C21 (step C22).
As a result, poor quality learning data is removed and good quality learning data is extracted.

Next, the learning unit 206 performs machine learning for each type of control using the learning data remaining in step C22. Here, it is assumed that a hierarchical neural network having M (M is a positive integer) layers is used as the machine learning algorithm (step C23).

First, the input signal of the neural network is expressed by the following equation 1.

Here, N represents the number of judgment information used for determining the control information. Here, it is assumed that the determination information acquired in the latest N seconds of the control date and time is used for determining the control information, and the determination information is acquired every second. Therefore, the learning data consists of N pieces of judgment information and one control information. In Equation 1, the value representing the N current bit rates acquired in N seconds is represented by BR, and the value representing the N current loop bands is represented by LB. The representative value may be the latest value, the average value, the maximum value, the minimum value, or the like among the N values, but may be other statistical values. Further, the N AGC reception levels acquired in N seconds are represented as LV = {LV (1), ..., LV (N)}. Further, N meteorological information acquired in N seconds is represented as WI = {WI (1), ..., WI (N)}. Further, the information representing the N weather forecast information acquired in N seconds is expressed as FC = {FC (1), ..., FC (4)}. Since one weather forecast information consists of four numerical values, here, for example, four values FC (1) to FC (4) of the latest weather forecast information are adopted as information representing N weather forecast information. Has been done. However, the information representing the N weather forecast responses may be other statistical information.

Further, when the output of the jth neuron in the mth layer of the neural network is expressed as o (m, j), the output of the first layer is expressed by the following equation 2 using the input signal.

The output of the other layer is expressed by the following equation 3.

Here, w (j, i, m-1) is a weight that is output from the i-th neuron in the m-1 layer and multiplied by the value input to the j-th neuron in the m-th layer. Represents. Further, L _m-1 represents the number of neurons in the m-1 layer. Further, f (x) represents an activation function, and for example, the following equation 4 is used.

Further, as a learning method of the neural network, for example, an error back propagation method is used. In the error back propagation method, as shown in the following equation 5, the weight w is calculated so as to minimize the square error between the output of the output layer and the teacher signal.

Here, y (i) represents a teacher signal. For example, consider machine learning using control information related to a bit rate set value as a teacher signal. Further, for example, it is assumed that there are three values that can be set as the bit rate: 50, 100, and 150. In this case, the set value 50 is assigned to y (1), the set value 100 is assigned to y (2), and the set value 150 is assigned to y (3). Here, when the bit rate setting value of the control information of the learning data is 50, y (1) = 1, y (2) = 0, y (3) = 0 is applied as the teacher signal. That is, y (i) represents whether or not the assigned set value is the set value of the teacher signal by 1 or 0.

In this way, the learning unit 206 individually constructs a neural network for bit rate setting control, loop band setting control, and operation end control, and calculates weights by the inverse error propagation method. Then, the learning unit 206 calculates the weight for the bit rate setting control, the weight for the loop band setting control, and the weight for the operation end control, respectively.

That is, the learning unit 206 performs weight learning of the first neural network using the judgment information of the learning data as an input signal and the bit rate set value included in the control information of the learning data as a teacher signal. Further, the learning unit 206 performs weight learning of the second neural network using the judgment information of the learning data as an input signal and the loop band set value included in the control information of the learning data as a teacher signal. Further, the learning unit 206 performs weight learning of the third neural network using the judgment information of the learning data as an input signal and the operation end control value included in the control information of the learning data as a teacher signal.

Next, the learning unit 206 stores the weights obtained for each type of control information as learning results in the learning result storage unit 107 (step C24).

This completes the description of the operation in which the information processing system 2 learns in this specific example.

Next, the operation of the information processing system 2 to calculate the control information will be described using the neural network for each type of control constructed in this way. Here, the calculation of the control information in the third operation mode will be described as an example.

Here, first, the control information calculation unit 208 reads the judgment information added in the last N seconds up to the current time from the judgment information storage unit 102 shown in FIG. 18 (step D31).

Next, the control information calculation unit 208 uses the read determination information as the input signal shown in Equation 1. Then, the control information calculation unit 208 reads the weight of the neural network constructed for the bit rate setting control from the learning result storage unit 107, and repeatedly applies Equations 2 and 3, respectively, to calculate the output of the output layer. do.

Then, the control information calculation unit 208 selects the control assigned to the neuron that outputs the value closest to 1 among the neurons in the output layer. For example, as shown in FIG. 22, the neural network output layer (third layer) of the bit rate setting control has three neurons of o (M, 1), o (M, 2), and o (M, 3). It is assumed that it is configured. Further, there are three values that can be set as the bit rate, 50, 100, and 150, and 50 is assigned to o (M, 1), 100 is assigned to o (M, 2), and 150 is assigned to o (M, 3). Suppose you are. At this time, it is assumed that the control information calculation unit 208 calculates o (M, 1) = 0.24, o (M, 2) = 0.89, and o (M, 3) = 0.10. In this case, the control information calculation unit 208 selects the bit rate setting value 100 assigned to o (M, 2) = 0.89, which is the closest to 1, as the control information for setting the bit rate (Tep D32).

Then, the control information calculation unit 208 transmits the bit rate set value 100 to the ground station control unit 105.

Then, the ground station control unit 105 determines whether or not the received bit rate setting value 100 is the content for changing the setting of the current transmission / reception equipment. When the content is to be changed (Yes in step D33), the ground station control unit 105 transmits the bit rate set value 100 to the ground station (step D34).

Similarly, the information processing system 2 also executes steps D32 to D34 for the loop band setting control and the operation end control.

Then, when the control information calculated for the operation end control is not the end of operation (No in step D35), the information processing system 2 repeats the operation from step D31.

On the other hand, when the control information calculated for the operation end control is the end of operation (Yes in step D35), the information processing system 2 ends the operation.

As described above, in this specific example, the neural network applied as machine learning is individually constructed for each type of control. Also, the number of neurons in the output layer of the neural network is set to be the same as the number of control values that can be specified in that type of control. As a result, the specific example of this embodiment enhances the calculation accuracy of the control information.

In order to compare with a specific example of this embodiment, a case where a neural network is not constructed for each type of control as machine learning will be described. In this case, for example, as shown in FIG. 23, a bit rate setting value is assigned to o (M, 1) of the output layer, a loop band setting value is assigned to o (M, 2), and O (M, 3) is assigned. It is assumed that an operation end control value (an arbitrary real value from 0 to 1 indicating the necessity of operation end) is assigned. As the operation end control value, when a threshold value (for example, 0.5) or more is calculated, 1 indicating the operation end is adopted as control information. At this time, it is assumed that o (M, 1) = 256, o (M, 2) = 100, and O (M, 3) = 0.8 are output. In this case, the control information in which the bit rate set value is 256, the loop band set value is 100, and the operation end control value is 1, is calculated. In such machine learning, one neural network must be used to learn weights for different types of control. Therefore, the calculation accuracy is lowered as compared with the case where the neural network is individually constructed for each type of control as in the specific example of the present embodiment.

Further, in order to compare with the specific example of this embodiment, even when a neural network is constructed for each type of control as machine learning, as shown in FIG. 24, one neuron in the output layer is used as one. A case where the output value is applied as a setting value will be described. However, when a value output from one neuron in the output layer is adopted as a set value, the values that can be calculated as the set value are various, and the calculation accuracy of the control information is lowered by that amount. On the other hand, in the specific example of this embodiment, the number of neurons in the output layer is set to the same number as the number of set values that can be specified. Thereby, in this embodiment, the output value of each neuron can be compared, and even if there is a slight error in the output value of each neuron, it is possible to absorb the error and select a more appropriate setting value. The sex is high.

This is the end of the explanation of the specific example.

Next, the effect of the second embodiment of the present invention will be described.

The information processing system as the present embodiment can automate the control of the device that controls the object with higher accuracy without the need to determine the control rule in advance.

The reason will be explained. In this embodiment, in addition to the same configuration as that of the first embodiment of the present invention, the evaluation value acquisition unit acquires the evaluation value for the control information and stores it in the evaluation value storage unit. This is because the learning unit performs machine learning using the learning data including the control information that is the evaluation target of the evaluation value satisfying the predetermined condition.

As a result, in the present embodiment, machine learning is performed by excluding the control information having a low evaluation from the control information based on the judgment by the operator and the control information calculated by the control information calculation unit from the learning target. be able to. As a result, the present embodiment can further improve the calculation accuracy of the control information. Further, in this embodiment, by using the evaluation value, it is possible to evaluate the performance of automating the calculation of control information while operating, and as a result, the human cost of engineers and the like involved in the evaluation. Is greatly reduced.

Further, in the present embodiment, the learning unit performs machine learning for each type of control represented by the control information, and the control information calculation unit applies the learning result for each type of control to the judgment information. This is because the control information is calculated and output.

As a result, the present embodiment can improve the learning accuracy as compared with the case of learning to derive different types of control information by one machine learning, and as a result, the calculation accuracy of the control information can be improved. ..

Further, in the present embodiment, the operation mode switching unit switches to any of the first, second, and third operation modes for operation. The first operation mode is an operation mode in which the control information acquisition unit and the learning unit function but the control information calculation unit does not function. The second operation mode is an operation mode in which the control information acquisition unit, the learning unit, and the control information calculation unit function together. Further, the third operation mode is an operation mode in which the control information calculation unit functions without the control information acquisition unit and the learning unit functioning.

In this way, in this embodiment, by providing the first to third operation modes in a switchable manner, it is possible to gradually shift from the operation by the operator to the operation by automation in a short period of time. do.

In the specific example of this embodiment, an example in which the control target is an artificial satellite has been described, but the control target is not limited to the artificial satellite. For example, the object to be controlled may be an aircraft that controls using an aeronautical radar.

Further, in a specific example of the present embodiment, an example in which the current bit rate, the current loop band, the weather information, and the weather forecast information are adopted as the judgment information has been described. Not limited to this, various other information representing the state of the ground station and the environment around the ground station can be applied as the judgment information.

For example, in the specific example of the present embodiment, one of the determination information may be information representing the environment around the antenna that communicates with the controlled object. Further, as ancillary equipment for detecting the environment around the antenna, a camera for monitoring the vicinity of the antenna may be applied. In this case, the present embodiment is based on not only the weather information but also the environment such as birds flying around the antenna as the environmental information, and the operation is terminated because the radio wave quality is poor. It is possible to learn control and use it for automation.

Further, for example, in the specific example of the present embodiment, one of the determination information may be information representing the azimuth angle and the elevation angle of the antenna. In this case, this embodiment can be used for automation by learning the control by the operator considering the direction in which the antenna is facing as the state of the ground station.

Further, in the specific example of this embodiment, the types of meteorological information shown in FIG. 19 are not limited to these. Also, this information does not necessarily have to be expressed numerically.

Further, in a specific example of the present embodiment, as shown in FIG. 20, an example in which three types of control, loop band setting control, bit rate setting control, and operation end control, are applied has been described. The types of are not limited to these.

Further, in the specific example of the present embodiment, it is exemplified that the value is equal to or higher than the threshold value as a predetermined condition of the evaluation value to be adopted as the learning data. Not limited to this, the predetermined condition of the evaluation value may be another condition or a combination of a plurality of conditions.

Further, in each of the above-described embodiments of the present invention, an example in which each functional block of the information processing system is distributed and realized in each device of the ground station control device, the automatic operation device, and the operation terminal has been described. Not limited to this, each functional block of the information processing system may be distributed and realized on a plurality of devices having any other configuration. Further, each functional block of the information processing system may be realized on one computer device.

Further, in each of the above-described embodiments of the present invention, an example in which each functional block of the information processing system is realized by a CPU that executes a computer program stored in a memory has been mainly described. Not limited to this, a part, all, or a combination thereof of each functional block may be realized by dedicated hardware.

Further, in each embodiment of the present invention described above, the operation of the information processing system described with reference to each flowchart is stored in the storage device (storage medium) of the computer device as the computer program of the present invention. Then, the CPU may read and execute the computer program. In such a case, the present invention is composed of the code or storage medium of the computer program.

Moreover, each of the above-described embodiments can be implemented in combination as appropriate.

Further, the present invention is not limited to each of the above-described embodiments, and can be implemented in various embodiments.

1, 2 Information processing system 101 Judgment information acquisition unit 102 Judgment information storage unit 103, 203 Control information acquisition unit 104 Control information storage unit 105 Ground station control unit 106, 206 Learning unit 107 Learning result storage unit 108, 208 Control information calculation unit 209 Evaluation value acquisition unit 210 Evaluation value storage unit 211 Operation mode switching unit 300 Ground station control device 400, 410 Automatic operation device 500, 510 Operation terminal 3001, 4001, 5001 CPU
3002, 4002, 5002 Memory 5003 Output device 5004 Input device 3005, 4005, 5005 Network interface

Claims (9)

A judgment information acquisition unit that acquires judgment information including information indicating the state of a ground station that communicates with a controlled object and information indicating the environment around the ground station.
A judgment information storage unit that stores the judgment information,
A control information acquisition unit that acquires control information indicating control contents for the ground station determined based on the determination information, and a control information acquisition unit.
A control information storage unit that stores the control information,
A ground station control unit that controls the ground station by transmitting the control information to the ground station,
A learning unit that performs machine learning using the judgment information stored in the judgment information storage unit and the control information stored in the control information storage unit as learning data.
A learning result storage unit that stores the learning results of the machine learning,
A control information calculation unit that calculates the control information by applying the learning result stored in the learning result storage unit to the determination information.
Equipped with
The learning unit performs the machine learning for each type of control represented by the control information, and performs the machine learning.
The control information calculation unit calculates the control information by applying the learning result for each type of control to the determination information.
The control information acquisition unit presents the determination information and the control information calculated by the control information calculation unit to the operator, and acquires the control information input by the operator.
The ground station control unit is an information processing system that controls the ground station by transmitting the control information input by the operator to the ground station. The control information calculation unit transmits the calculated control information to the ground station control unit.
The information processing system according to claim 1, wherein the ground station control unit transmits the control information received from the control information calculation unit to the ground station. The control information calculation unit transmits the calculated control information to the control information acquisition unit.
The first or second aspect of the present invention, wherein the control information acquisition unit acquires the control information determined based on the control information received from the control information calculation unit and the determination information. Information processing system. An evaluation value acquisition unit that acquires an evaluation value representing an evaluation of the control information, and an evaluation value acquisition unit.
An evaluation value storage unit that stores the evaluation value and
Further prepare
The learning unit is claimed from claim 1, wherein the learning unit performs the machine learning using the learning data including the control information which is the evaluation target of the evaluation value satisfying a predetermined condition among the learning data. Item 3. The information processing system according to any one of items 3. A first operation mode in which the control information acquisition unit and the learning unit function but the control information calculation unit does not function, and a second operation in which the control information acquisition unit, the learning unit, and the control information calculation unit both function. Further provided with a mode and an operation mode switching unit for switching to any of the third operation modes in which the control information acquisition unit and the learning unit do not function and the control information calculation unit functions.
In the second operation mode, the control information acquisition unit acquires the control information input by the operator based on the determination information and the control information calculated by the control information calculation unit.
In the second operation mode, the learning unit further determines a combination of the determination information and the control information input by the operator or the control information calculated by the control information calculation unit. The machine learning is performed using a combination that satisfies the conditions as learning data.
The information processing system according to any one of claims 1 to 4, wherein the information processing system is characterized by the above. In the information processing system according to any one of claims 1 to 5.
An automatic operation device having the determination information storage unit, the control information storage unit, the learning unit, the learning result storage unit, and the control information calculation unit. In the information processing system according to any one of claims 1 to 5.
An operation terminal having the control information acquisition unit. Judgment information including information indicating the state of the ground station communicating with the controlled object and information representing the environment around the ground station, and the control content for the ground station determined based on the judgment information. When performing machine learning using control information as learning data, the machine learning is performed for each type of control represented by the control information.
The control information is calculated by applying the learning result of the machine learning for each type of control to the judgment information.
A method of presenting the determination information and the calculated control information to an operator, and transmitting the control information input by the operator to the ground station to control the ground station. Judgment information including information indicating the state of the ground station communicating with the controlled object and information representing the environment around the ground station, and the control content for the ground station determined based on the judgment information. When performing machine learning using control information as learning data, the machine learning step for performing the machine learning for each type of control represented by the control information, and
A control information calculation step for calculating the control information by applying the learning result of the machine learning for each type of control to the determination information.
A control step in which the determination information and the calculated control information are presented to the operator, and the control information input by the operator is transmitted to the ground station to control the ground station.
A program that causes a computer device to execute.

Images