WO2023170890A1

WO2023170890A1 - Specifying device, program, learning method, learning device, and trained model

Info

Publication number: WO2023170890A1
Application number: PCT/JP2022/010713
Authority: WO
Inventors: 誠野村
Original assignee: 三共木工株式会社
Priority date: 2022-03-10
Filing date: 2022-03-10
Publication date: 2023-09-14

Abstract

According to the present invention, an aircraft response characteristic providing system comprises: an aircraft holding device that holds a mounted drone; and an information output device that displays information for specifying data related to a failure of the drone mounted on the aircraft holding device. The drone, while held by the aircraft holding device, receives an instruction signal indicating the content of movement from the operating device and moves in response to the received instruction signal. The information output device specifies data regarding the failure of the drone, on the basis of an ideal movement state of the drone and an actual movement state of the drone in response to the instruction signal, and a trained model.

Description

Specific equipment, program, learning method, learning device, and trained model

The technology of the present disclosure relates to a specific device, a program, a learning method, a learning device, and a model.

The fuel cell-equipped drone disclosed in Japanese Patent Application Laid-open No. 2019-145381 is capable of operating each electric motor using the power supplied from the other fuel cells even if the output of one of the first to third fuel cells decreases. By driving the aircraft, it prevents the aircraft from being unable to fly or crashing due to a drop in fuel cell output.

However, the cause of the decrease in the output of one of the first to third fuel cells has not been identified.

The technology of the present disclosure includes a specific device and program that can specify data related to aircraft performance, a learning method that generates a learned model by learning a model that specifies data related to aircraft performance, a learning device, and a program, and a trained model for identifying data regarding aircraft performance.

In order to achieve the above object, the identification device according to the first aspect of the technology of the present disclosure receives a signal indicating the content of the movement from the instruction device, and identifies data regarding the performance of the moving aircraft according to the received signal. a receiving unit that receives the signal, and identifies data regarding the performance of the aircraft based on an ideal movement state of the aircraft and an actual movement state of the aircraft according to the received signal. and a specific part.

In the first aspect, the identification device according to a second aspect of the technology of the present disclosure stores a plurality of combinations of an ideal movement state of the aircraft, an actual movement state of the aircraft, and data regarding the performance of the aircraft. The determination unit further includes a storage unit that performs a determination based on the plurality of combinations stored in the storage unit, an ideal movement state of the aircraft according to the received signal, and an actual movement state of the aircraft. and identifying data regarding the performance of the aircraft.

In the first aspect, the identification device according to a third aspect of the technology of the present disclosure is a teacher that receives an ideal movement state of the aircraft and an actual movement state of the aircraft as input, and outputs data regarding the performance of the aircraft. further comprising a storage unit that stores a trained model trained using the data,
The determination unit determines the performance of the aircraft based on the learned model stored in the storage unit, an ideal movement state of the aircraft and an actual movement state of the aircraft according to the received signal. Identify data.

The identification device according to a fourth aspect of the technology of the present disclosure is such that, in any one of the first to third aspects, at least one of the aircraft and the identification device determines the actual movement state of the aircraft. The actual movement state of the aircraft used by the determination unit is the actual movement state detected by the detection unit.

In the identification device according to a fifth aspect of the technology of the present disclosure, in any one of the first to fourth aspects, the aircraft is held movably in three dimensions by an aircraft holding device. move in a state.

In the identification device according to a sixth aspect of the technology of the present disclosure, in any one of the first to fifth aspects, the data regarding the performance of the aircraft is designed to have a predetermined response characteristic. data relating to improvements in the aircraft during the current stage of development or failures of the aircraft manufactured according to a completed design with predetermined response characteristics.

A program according to a seventh aspect of the technology of the present disclosure is a learned program used for receiving a signal indicating the content of movement from an instruction device and identifying data regarding the performance of a moving aircraft in response to the received signal. A program that causes a computer to execute a specific process that uses a model to specify data regarding the performance of an aircraft, the specific process including an ideal movement state of the aircraft in response to the signal and an actual movement state of the aircraft. and identifying data regarding the performance of the aircraft based on the trained model, the acquired ideal movement state of the aircraft, and the actual movement state of the aircraft.

A learning method according to an eighth aspect of the technology of the present disclosure is a learning method that generates a trained model by learning a model that specifies data related to the performance of an aircraft, the learning method specifying data related to the performance of the aircraft. a step of obtaining training data having input information for identifying the performance of the aircraft and outputting data regarding the performance of the aircraft; using the training data, inputting information for specifying data regarding the performance of the aircraft; training a model whose output is data regarding aircraft performance.

A learning device according to a ninth aspect of the technology of the present disclosure is a learning device that generates a learned model by learning a model that specifies data related to the performance of an aircraft, the learning device specifying data related to the performance of the aircraft. an acquisition unit that acquires training data having input information for determining the performance of the aircraft and outputting data regarding the performance of the aircraft, and inputting information for specifying data regarding the performance of the aircraft using the training data; and a learning processing unit that learns a model that outputs data regarding the performance of the aircraft.

A program according to a tenth aspect of the technology of the present disclosure is a program that causes a computer to execute a learning process of generating a learned model by learning a model that specifies data related to the performance of an aircraft, the program The process includes a step of obtaining training data having input information for specifying data regarding the performance of the aircraft and outputting data regarding the performance of the aircraft, and using the training data to obtain data regarding the performance of the aircraft. The method includes the step of learning a model that inputs information for specifying the aircraft and outputs data regarding the performance of the aircraft.

The learned model according to the eleventh aspect of the technology of the present disclosure is a learned model for specifying data related to the performance of an aircraft, and the learned model includes information for specifying data related to the performance of the aircraft. The trained model receives as input information for specifying data regarding the performance of the aircraft, and receives as input information for specifying data regarding the performance of the aircraft. A computer is caused to execute a process for specifying data related to the performance of the aircraft based on information for specifying data related to the performance of the aircraft.

The first aspect of the technology of the present disclosure can identify data regarding aircraft performance.

The second aspect can eliminate the need for model learning.

The third aspect is that, compared to specifying data regarding aircraft performance using a plurality of combinations stored in the storage unit, data regarding aircraft performance can be specified even for combinations that are not stored. .

A fourth aspect may use the actual movement state of the aircraft detected by a detection unit provided in at least one of the aircraft and the specific device.

The fifth aspect can prevent the aircraft from crashing.

The sixth aspect is data regarding the performance of the aircraft, including data regarding improvements to the aircraft at the stage of designing it to have predetermined response characteristics, or data regarding improvements to the aircraft at the stage of designing the aircraft to have predetermined response characteristics, or data regarding the improvement of the aircraft in accordance with a completed design to have predetermined response characteristics. data relating to failures of the aircraft may be identified.

A seventh aspect is that data regarding the performance of the aircraft can be specified.

An eighth aspect can provide a learning method that can generate a model for specifying data regarding aircraft performance.

A ninth aspect can provide a learning device that can generate a model for specifying data regarding aircraft performance.

A tenth aspect can provide a program that can generate a model for specifying data regarding aircraft performance.

An eleventh aspect can provide a trained model for identifying data regarding aircraft performance.

1 is a schematic diagram of an aircraft response characteristic providing system 100 according to an embodiment. 2 is a diagram mainly showing the schematic configurations of a drone 10 and an aircraft holding device 150. FIG. FIG. 3 shows the aircraft holding device 150 with the extendable strut 110 extended. FIG. 7 is a cross-sectional view showing the structure of the upper end portion of the cylindrical body 110N1 and the lower end portion of the cylindrical body 110N2 in a state where the support column 110 is extended. This is a diagram showing how the mounting table 114 is tilted downward on the right side and upward on the left side when viewed from the paper, with the drone 10 placed on the mounting table 114 and the support column 110 extended. be. 2 is a schematic diagram of an information output device 170. FIG. 2 is a functional block diagram of a CPU 212 of the information output device 170. FIG. It is a flowchart of the information output processing program 222P1 executed by the CPU 212 of the information output device 170. It is a flowchart of the learning processing program 222P2 executed by the CPU 212 of the information output device 170. It is a flowchart of the identification program 222P3 of the first process for identifying data related to a failure, which is executed by the CPU 212 of the information output device 170. FIG. 22 is a diagram showing a screen 224S of the display 224 that displays information for specifying data regarding a malfunction of the drone 10. FIG. It is a graph of the change in height of the drone 10 over time, displayed on the display 224, from when the start button is turned on until when the stop button is turned on. It is a graph of the change in pitch angle of the drone 10 over time from when the start button is turned on until the stop button is turned on, displayed on the display 224. It is a flowchart of the identification program of the 2nd process which automatically identifies the data regarding the failure of the drone 10, which is executed by the CPU 212 of the information output device 170. It is a figure showing the schematic structure of aircraft holding device 150C1 of the 1st modification. 7 is a diagram illustrating how the drone 10 of the first modification example rises while being held on the mounting table 114. FIG. It is a figure showing the schematic structure of aircraft holding device 150C2 of the 2nd modification. 10 is a diagram showing a state in which the drone 10 of the second modification example has slightly risen from the state held on the mounting table 114, and the support columns 420N1 to 420N4 have fallen down. FIG.

Hereinafter, embodiments of the technology of the present disclosure will be described with reference to the drawings.

(First embodiment)
FIG. 1 shows a schematic diagram of an aircraft response characteristics providing system 100 according to an embodiment. As shown in FIG. 1, the aircraft response characteristic providing system 100 includes an aircraft holding device 150 (see also FIG. 2) that holds the aircraft 10 mounted thereon, and a response characteristic of the aircraft 10 mounted on the aircraft holding device 150. and an information output device 170 that outputs (for example, displays) information for evaluating. The aircraft response characteristics providing system 100 is placed indoors where there is no possibility of receiving crosswinds or the like. Therefore, the limited flight test described below can be conducted in windless conditions.
The aircraft response characteristics providing system 100 is an example of a "specific system" of the technology of the present disclosure. The information output device 170 is an example of a “specific device” and a “learning device” of the technology of the present disclosure.

In this embodiment, a drone 10 will be described as an example of an aircraft.

The operator operates the operating device 50 in order to move the drone 10 in a desired direction (that is, fly). The operating device 50 transmits to the drone 10 an instruction signal indicating the content of movement according to the content of the operation. The drone 10 has a function of moving in accordance with an instruction signal received from the operating device 50 and indicating the content of movement. Drone 10 is manufactured according to a completed design with predetermined response characteristics.

The operating device 50 is an example of the "instruction device" of the technology of the present disclosure.

FIG. 2 mainly shows the schematic configurations of the drone 10 and the aircraft holding device 150. Since the configuration of the drone 10 is well known, a detailed explanation will be omitted, but as shown in FIG. , a propeller 18 rotated by the motor 16, and four support parts 20 that support the main body 12. The main body 12 includes a communication device and a flight controller (not shown). When the communication device receives an instruction signal indicating the content of the movement from the operating device 50, the flight controller controls each motor 16 so that the drone 10 moves according to the content of the movement indicated by the received instruction signal. .

The aircraft holding device 150 includes a mounting table 114, a holding part 116 that holds the drone 10 placed on the mounting table 114, and a column 110 that supports the mounting table 114 at one end (i.e., the upper end) and is expandable and retractable in the vertical direction. , a universal joint 112 that connects one end of the support 110 and the support 114 such that the support 114 is three-dimensionally rotatable with respect to one end of the support 110. As shown in FIG. 1, the mounting table 114 has a plurality of mesh-like openings formed therein for weight reduction. As the holding part 116, for example, a binding band or the like with which the operator ties the supporting part 20 of the drone 10 to the mounting table 114 can be used.

The universal joint 112 is an example of a "connection part" of the technology of the present disclosure. Note that a ball joint may be used instead of the universal joint.

The aircraft holding device 150 includes a base body 118, a plurality of support columns 120N1 to 120N4 provided on the upper surface of the base body 118, and a plurality of casters (moving members) 122N1 to 122N4 provided on the lower surface of the base body 118. The shape and size of the upper surface of the base 118 are approximately the same as the shape and size of the lower surface of the mounting table 114, and the shape of the upper surface of the base 118 and the shape of the lower surface of the mounting table 114 are, for example, square. The support columns 120N1 to 120N4 are fixed to each corner of the upper surface of the base body 118. The number of support columns 120N1 to 120N4 is four.

The support columns 120N1 to 120N4 support the mounting table 114 at one end (that is, the upper end), and have their lower ends fixed to the upper surface of the base 118.

The casters 122N1 to 122N4 are examples of "moving members" of the technology of the present disclosure.

Since the aircraft holding device 150 is provided with a plurality of casters 122N1 to 122N4 on the lower surface of the base body 118, it can move along the plane where the aircraft holding device 150 is provided.

FIG. 3 shows the expandable support column 110 in an extended state. As shown in FIG. 3, the support column 110 includes a plurality of (for example, four) cylindrical bodies 110N1 to 110N4, each having the same length in the axial direction and different cross-sectional radii. The plurality of cylindrical bodies 110N1 to 110N4 are arranged concentrically from the outside to the inside in descending order of cross-sectional radius. The three inner cylindrical bodies 110N2 to 110N4 are movable in the vertical direction (that is, the axial direction), but the lower end of the outermost cylindrical body 110N1 is fixed to the base body 118.

FIG. 4 shows a cross-sectional view showing the structure of the upper end of the outer cylindrical body 110N1 and the lower end of the inner cylindrical body 110N2 in a state where the support 110 is extended. As shown in FIG. 4, a groove is formed in the axial direction from the upper end to the lower end on the inner surface of the cylindrical body 110N1. (For example, four) rollers 110N11 are rotatably attached around a shaft 110N12. In this way, the outer cylindrical body 110N1 contacts the inner cylindrical body 110N2 via the plurality of rotatably attached rollers 110N11. Therefore, the friction coefficient between the cylindrical body 110N1 and the cylindrical body 110N2 when the cylindrical body 110N2 ascends and descends is determined by the friction coefficient between the cylindrical body 110N2 and the outer surface of the cylindrical body 110N1 without providing the roller 110N11. can be made smaller than when they rise and fall in direct contact with each other. Although not shown, at the lower end of the cylindrical body 110N2, in order to prevent the cylindrical body 110N2 from coming off the cylindrical body 110N1 when the cylindrical body 110N2 rises, there is at least one At least one protrusion is provided at a position that contacts the roller 110N11. In this embodiment, four protrusions are provided, and each of the four protrusions is arranged at a position corresponding to each of the four rollers 110N11. The rollers 110N11 are provided in the cylindrical bodies 110N2 and 110N3 as well, and the cylindrical bodies 110N3 and 110N4 are provided with protrusions in the same manner, so a description thereof will be omitted.
Note that a ball (ie, a sphere) may be used instead of the roller.

As the mounting table 114 rises as the aircraft 10 rises, the support column 110 extends. Specifically, as the mounting table 114 rises, the cylindrical body 110N4, which is connected to the mounting table 114 via the universal joint 112, is pulled by the mounting table 114 and rises. As the cylindrical body 110N4 continues to rise, each protrusion of the cylindrical body 110N4 hits the roller 110N11 of the cylindrical body 110N3, and the cylindrical body 110N3 begins to rise. As the cylindrical body 110N3 continues to rise, each protrusion of the cylindrical body 110N3 hits the roller 110N11 of the cylindrical body 110N2, and the cylindrical body 110N2 begins to rise. As the cylindrical body 110N2 continues to rise, each protrusion of the cylindrical body 110N2 hits the roller 110N11 of the cylindrical body 110N1. The lower end of the cylindrical body 110N1 is fixed to the base body 118. Therefore, when the cylindrical body 110N2 rises until each protrusion of the cylindrical body 110N2 hits the roller 110N11 of the cylindrical body 110N1, each of the cylindrical bodies 110N4 to 110N2 stops rising. In this case, the length of the support column 110 is a predetermined maximum length.

When the mounting table 114 descends as the aircraft 10 descends, the cylindrical bodies 110N4 to 110N2 descend. When the lower end of the cylindrical body 110N2 reaches the base 118, the descent of the cylindrical body 110N2 is stopped, and the cylindrical bodies 110N4 and 110N3 are lowered. When the lower end of the cylindrical body 110N3 reaches the base 118, the descent of the cylindrical body 110N3 stops, and the cylindrical body 110N4 descends. When the lower end of the cylindrical body 110N4 reaches the base 118, the descent of the cylindrical body 110N4 is stopped. In this case, the length of the strut 110 is a predetermined minimum length.

In FIG. 5, the drone 10 is held on the mounting table 114 and the support column 110 is extended, and the mounting table 114 rotates three-dimensionally with respect to one end of the support column 110, specifically, toward the paper surface. The right side is shown tilting downward and the left side tilting upward. As described above, one end of the support column 110 and the mounting table 114 are connected by the universal joint 112. Therefore, the mounting table 114 can rotate three-dimensionally based on one end of the support column 110. Further, as described above, the aircraft holding device 150 can be moved along the plane where the aircraft holding device 150 is provided by the plurality of casters 122N1 to 122N4.

As described above, the drone 10 moves in response to an instruction signal indicating the content of movement received from the operating device 50. Specifically, the drone 10 moves in each three-dimensional direction, for example, upward movement, downward movement, horizontal direction, a combination of upward movement and horizontal direction (i.e. diagonally upward direction), and a combination of downward movement and horizontal direction. It has the function of being able to move in the opposite direction (that is, diagonally downward).

As described above, the drone 10 is held by the holding part 116 on the mounting table 114 which is fixed to the base body 118 and connected to one end of the extensible support column 110.

Therefore, for example, when the drone 10 ascends, it ascends together with the mounting table 114 until the support 110 reaches its maximum length. When the drone 10 descends, the drone 10 descends together with the mounting table 114 until the support 110 reaches its minimum length.

Furthermore, when the drone 10 attempts to move horizontally, the drone 10 moves horizontally along with the mounting table 114, the strut 110, and the base 118, that is, the entire aircraft holding device 150 moves horizontally. .

Further, when moving in a direction that combines upward and horizontal directions (that is, diagonally upward), the drone 10 ascends while tilting until the support column 110 reaches its maximum length, and the entire aircraft holding device 150 moves horizontally. Note that when moving diagonally upward, the inclination angle of the drone 10 is smaller than when simply moving horizontally, and the amount of movement per unit time that the entire aircraft holding device 150 moves in the horizontal direction is smaller. It's small.

Furthermore, when moving in a direction that combines descending and horizontal directions (that is, diagonally downward), the drone 10 descends while tilting until the support column 110 reaches its minimum length, and the entire aircraft holding device 150 moves horizontally. Note that when moving diagonally downward, the inclination angle of the drone 10 is smaller than when simply moving horizontally, and the amount of movement per unit time that the entire aircraft holding device 150 moves in the horizontal direction is smaller. It's small.

Then, if there is no problem with the drone 10 and the drone 10 has predetermined response characteristics, the drone 10 moves in accordance with the instruction signal indicating the content of movement received from the operating device 50. Here, the response characteristic of the technology of the present disclosure refers to a characteristic indicating how far the drone 10 can move according to the content of movement indicated by the instruction signal from the operating device 50. Therefore, for example, if the drone 10 has a predetermined response characteristic, the drone 10 will rise when instructed to ascend by the operating device 50, descend when instructed to descend, and descend when instructed to move horizontally. If instructed to move diagonally upward or downward, it moves diagonally upward or downward.

FIG. 6A shows a schematic block diagram of the information output device 170. As shown in FIG. 6A, the information output device 170 includes a plurality of cameras 202N1 to 202N4 (see also FIG. 1) that detect the movement state of the drone 10 from different directions, and a calculation storage device 200. . The cameras 202N1 to 202N4 have a communication function to communicate images obtained by shooting. The cameras 202N1 to 202N4 are placed on the indoor wall or pillar.
Cameras 202N1 to 202N4 are examples of "detection units" of the technology of the present disclosure.

The cameras 202N1 to 202N4 are arranged to photograph the moving state of the drone 10 from four different directions. Specifically, the camera 202N1 and the camera 202N3 are arranged to face each other, and the camera 202N2 and the camera 202N4 are arranged to face each other. More specifically, the directions of the central axes of the camera 202N1 and the camera 202N3 for photographing the drone 10 are located on the first plane. The directions of the central axes of the camera 202N2 and the camera 202N4 for photographing the drone 10 are located on the second plane. The first plane and the second plane intersect at right angles.

The calculation storage device 200 includes a computer 210, a secondary storage device 222, a communication interface (I/F) 226, an input device 228, and a display 224, each connected to the computer 210. The communication interface (I/F) 226 receives an instruction signal indicating the content of the movement from the operating device 50, and also receives image signals from the plurality of cameras 202N1 to 202N4. The computer 210 includes a CPU (Central Processing Unit) 212, a ROM (Read Only Memory) 214, a RAM (Random Access Memory) 216, and an input/output (I/O) port 218. CPU 212, ROM 214, RAM 216, and I/O port 218 are interconnected via bus 220. A secondary storage device 222, a communication interface (I/F) 226, an input device 228, and a display 224 are connected to the I/O port 218. In the example shown in FIG. 6A, the computational storage device 200 includes one communication interface (I/F) 226. However, a plurality of communication interfaces (I/F) 226 may be provided for the operating device 50 and the plurality of cameras 202N1 to 202N4. The input device 228 is, for example, a mouse, a keyboard, or the like.
The communication interface (I/F) 226 is an example of a “receiving unit” of the technology of the present disclosure.

The secondary storage device 222 includes programs such as an information output processing program 222P1 (see FIG. 7A), a learning processing program 222P2 (see FIG. 7B), and a specific program 222P3 (see FIG. 8 or 12), which will be described later, and a model 222M. is remembered. Each program is read from the ROM 214 (or the secondary storage device 222) to the RAM 214 and executed by the CPU 212 to perform information output processing, learning processing (i.e., machine learning processing), and specific processing, which will be described later. Note that each of the above programs may be stored in the ROM 214. The secondary storage device 222 is a non-transitory tangible computer readable media, such as a hard disk drive (HDD) or solid state drive (SSD). non-volatile It is a storage device.
The specific program 222P3 is an example of "a program that causes a computer to execute specific processing" of the technology of the present disclosure. Learning processing program 222P2 is an example of "a program that causes a computer to execute learning processing" of the technology of the present disclosure. The secondary storage device 222 is an example of a "storage unit" and a "recording medium" in the technology of the present disclosure.

In the ROM 214 or the secondary storage device 222, response characteristics of each of a plurality of types of drones having different response characteristics are stored in correspondence with data indicating the type of drone. As described above, the response characteristic of the technology of the present disclosure refers to a characteristic indicating how far the drone can move according to the content of movement indicated by the instruction signal from the operating device. Movement includes the elements of rising, falling, horizontal movement, diagonally upward movement, and diagonally downward movement. Therefore, specifically, in the ROM 214 or the secondary storage device 222, the amount of movement per unit time of ascending or descending, the inclination angle of the drone 10, and the amount of movement per unit time of each type of drone with the mounting table 114 and according to the instruction signal. Further, the amount of movement of the drone 10 in the horizontal direction per unit time is stored. In this way, since the response characteristics of each type of drone are stored in the ROM 214 or the secondary storage device 222, the CPU 212 of the information output device 170 receives the instruction signal using the response characteristics of the type of drone 10. It is possible to estimate the position and attitude of the drone 10 after a predetermined period of time has passed.

The model 222M stored in the secondary storage device 222 is a learning model for identifying data related to a failure of the drone 10. The model 222M is, for example, a neural network (NN). Specifically, the model 222M includes an input layer in which the ideal movement state of the drone 10 and the actual movement state of the drone 10 are input, an output layer in which data regarding the failure of the drone 10 is output, and an ideal movement state of the drone 10. and one intermediate layer in which parameters are learned using a plurality of pieces of teacher data, the input of which is the movement state of the drone 10 and the actual movement state of the drone 10, and the output is data regarding a failure of the drone 10. The model 222M causes the computer 210 to acquire the ideal movement state of the drone 10 and the actual movement state of the drone 10, and input the obtained ideal movement state of the drone 10 and the actual movement state of the drone 10 to the input layer. The intermediate layer performs calculations, and the output layer outputs data related to the failure.
The model 222M is an example of a "model" of the technology of the present disclosure.

The model 222M may be a deep neural network (DNN). In this case, model 222M includes multiple intermediate layers.

FIG. 6B shows a functional block diagram of the CPU 212. The functions of the CPU 212 include a learning function, a processing function, an estimation function, an import function, a calculation function, a storage function, a reading function, a display function, and a specific function. The learning function includes an acquisition function and a learning processing function. As shown in FIG. 6B, by executing any of the programs described above, the CPU 212 includes a learning section 500, a processing section 502, an estimating section 504, an importing section 506, a calculating section 508, a storage section 510, a reading section 512, It functions as a display section 514 and a specifying section 516. The learning section 500 includes a learning processing section 500B.
The specifying unit 516 is an example of the “specific unit” of the technology of the present disclosure. The acquisition unit 500A is an example of the “acquisition unit” of the technology of the present disclosure. The learning processing unit 500B is an example of a “learning processing unit” of the technology of the present disclosure.

Next, the operation of this embodiment will be explained.
(Flight test)
First, although the details will be described later, in this embodiment, the information output device 170 acquires information for specifying data regarding a failure of the drone 10 by conducting a flight test.
(Model learning)
Second, although it will be described in detail later, in this embodiment, the information output device 170 uses information for identifying data related to a failure of the drone 10, which is obtained by performing such a flight test. , model 222M is trained.
(Identification of data regarding failure of drone 10)
Thirdly, in this embodiment, the information output device 170 uses the trained model 222M to specify data regarding a failure of the drone 10, although this will be described in detail later.

(Flight test)
First, I will explain the flight test. In this embodiment, a free flight test in which the drone 10 actually flies freely is not performed, but the operator performs a restricted flight test with the drone 10 held in the aircraft holding device 150 so as to be movable in three dimensions. conduct. This is for the following reason.

If the assembled drone 10 has predetermined response characteristics, the drone 10 moves in accordance with the instruction signal received from the operating device 50 and indicating the content of movement. However, if the assembled drone 10 has some kind of defect and does not have the predetermined response characteristics, the drone 10 will respond to the instruction signal indicating the content of movement received from the operating device 50. Don't move.

At the stage when the assembly is completed, it is not known whether the drone 10 has predetermined response characteristics. In order to confirm whether the assembled drone 10 has predetermined response characteristics, it is conceivable to perform a free flight test in which the drone 10 is actually allowed to fly freely.

However, if the drone 10 subjected to the free flight test does not have predetermined response characteristics, it may crash and be destroyed during movement.

Therefore, in this embodiment, in order to confirm whether the assembled drone 10 has predetermined response characteristics, a free flight test in which the drone 10 is actually flown freely is not performed, and the operator A limited flight test is performed in which the drone 10 is moved while being held movably in three dimensions by the aircraft holding device 150. Specifically, the operator first places the drone 10 on the mounting table 114 and causes the tips of the four supporting parts 20 to be held on the mounting table 114 by the holding parts 116. Next, the operator operates the operating device 50 in order to move the drone 10 held on the mounting table 114. The operated operating device 50 transmits an instruction signal indicating the content of the movement so as to move according to the operated content. The instruction signal indicating the content of movement transmitted from the operating device 50 is received by the drone 10, and the drone 10 attempts to move according to the received instruction signal indicating the content of movement. For example, as described above, when the drone 10 attempts to ascend or descend, it attempts to ascend or descend together with the mounting table 114. Furthermore, when the drone 10 attempts to move horizontally, the entire aircraft holding device 150 attempts to move horizontally. The information output device 170 also receives the instruction signal indicating the content of the movement transmitted from the operating device 50 .

FIG. 7A shows a flowchart of the information output processing program 222P1 executed by the CPU 212 of the information output device 170. The information output processing program 222P1 shown in FIG. 7A provides information for specifying data regarding a failure of the drone 10 so that it can be later confirmed whether the assembled drone 10 has predetermined response characteristics. is output (ie, stored) to the secondary storage device 222.

The information output processing program 222P1 starts when data indicating the type of the drone 10 is input from the input device 228 and a start button (not shown) connected to the computer 210 is turned on. Ends when the stop button is turned on. Alternatively, the communication I/F 226 may start when receiving a start instruction signal from the operating device 50 and end when receiving a stop instruction signal from the operating device 50.

When the start button is turned on, in step 602, the processing unit 502 determines whether an instruction signal indicating the movement of the drone 10 has been received from the operating device 50 via the communication I/F 226.

As described above, the operating device 50 operated by the operator transmits an instruction signal indicating the content of the movement so that the operating device 50 moves according to the content of the operation. The instruction signal indicating the content of movement transmitted from the operating device 50 is received by the drone 10 and also by the communication I/F 226. When the instruction signal indicating the content of movement is received by the communication I/F 226, the determination in step 602 becomes affirmative. If the instruction signal indicating the content of the movement is not received by the communication I/F 226, the determination in step 602 becomes a negative determination, and the process in step 602 is repeatedly executed until the determination in step 602 becomes an affirmative determination.

If the determination in step 602 is affirmative, the information output process proceeds to step 604. In step 604, the capture unit 506 captures each image signal from each of the cameras 202N1 to 202N4 via the communication I/F 226.

In step 606, the processing unit 502 determines whether time t has elapsed since each image signal was captured in step 604. If the determination in step 606 is negative, the process in step 606 is repeatedly executed until the determination in step 606 is positive.

If the determination in step 606 is affirmative, the information output process proceeds to step 608. In step 608, the capture unit 506 captures each image signal from each of the cameras 202N1 to 202N4 via the communication I/F 226.

In step 610, the calculation unit 508 calculates the position and orientation of the drone 10 from each image signal captured in step 604, and calculates the position and orientation of the drone 10 from each image signal captured in step 604 and each image signal captured in step 608. From this, the moving direction and moving speed of the drone 10 are calculated.

Specifically, the calculation unit 508 performs image matching or the like on each image signal captured in step 604 and each image signal captured in step 608 to find a predetermined position of the drone 10 in each image (for example, the main body). 12), and from the specified predetermined position, the position of the predetermined position in a three-dimensional space, which will be described later, is calculated using the principle of triangulation. The calculation unit 508 calculates the attitude of the drone 10 based on, for example, the degree of inclination of the main body 12 of the drone 10 photographed in advance by each of the cameras 202N1 to 202N4.

The calculation unit 508 calculates the moving direction and movement of the drone 10 from the position of the main body 12 in the image of each image signal captured in step 604 and the position of the main body 12 in the image of each image signal captured in step 608. Calculate speed. Note that time t is used when calculating the moving speed.

In step 612, the estimation unit 504 calculates the response characteristics stored in the ROM 214 or the secondary storage device 222 in correspondence with the data indicating the type of the drone 10, the contents of the instruction signal, the position, attitude, and movement of the drone 10. From the direction and movement speed, the position and attitude of the drone 10 after a predetermined period of time from the time when the determination in step 602 is affirmative is estimated. Note that the predetermined time is a predetermined time longer than the time t.

In step 614, the processing unit 502 determines whether a predetermined time has elapsed since the determination in step 602 was affirmative. If the determination at step 614 is negative, the process at step 614 is repeatedly executed until the determination at step 614 is affirmative.

If the determination in step 614 is affirmative, the information output process proceeds to step 616.

In step 616, the capture unit 506 captures the actual movement of the drone 10, for example, each image from each of the cameras 202N1 to 202N4, via the communication I/F 226.

In step 618, the calculation unit 508 calculates the position and attitude of the drone 10 from each captured image.

In step 620, the storage unit 510 stores the position and orientation of the drone 10 estimated in step 612 (that is, the ideal position and orientation (also referred to as ideal movement state)) and the position and orientation of the drone 10 calculated in step 618. The posture (that is, the actual position and posture (also referred to as the actual movement state)) is stored in the secondary storage device 222 along with each image captured in step 616 and the elapsed time since the information output processing program 222P was started. ,Output.

In step 622, the processing unit 502 determines whether to end the information output processing program 222P by determining whether a stop button (not shown) has been turned on. If the stop button is not turned on, a negative determination is made in step 622, and the information output process returns to step 602 to execute the above processes (ie, steps 602 to 622). When the stop button is turned on, an affirmative determination is made in step 622, and the information output processing program 222P ends.

As explained above, each time an instruction signal is received from the operating device 50 from when the start button is turned on until the stop button is turned on, the ideal position and attitude of the drone 10 and the actual The position and orientation are stored in the secondary storage device 222 in correspondence with each image captured in step 616 and the elapsed time since the information output processing program 222P1 was started. If the instruction signal is transmitted from the operating device 50 multiple times, for example, N times, between when the start button is turned on and when the stop button is turned on, the ideal position and attitude of the drone 10 and the actual position are transmitted. N combinations of the image and posture are stored in the secondary storage device 222 along with each image captured in step 616 and the elapsed time since the information output processing program 222P1 was started.

(Model learning)
FIG. 7B shows a flowchart of the learning process program 222P2 of the learning process in which the learning unit 500 in the CPU 212 of the information output device 170 learns the model 222M using teacher data. When the learning unit 500 in the CPU 212 of the information output device 170 executes the learning processing program 222P2, the learning method is executed and the learned model 222M is generated. The learning processing program 222P2 starts when a learning instruction button (not shown) is turned on.

At step 500SA, the acquisition unit 500A in the learning unit 500 acquires teacher data.

Here, the teaching data will be explained.
The display unit 514 controls the display 224 so that the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 are displayed (FIGS. 8 to 11). By looking at the display of the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10, and by looking at the flight condition of the drone 10 in the above-mentioned limited flight test, the operator can adjust the predetermined position and attitude to the drone 10. It can be determined whether or not the device has a response characteristic.

If the operator determines that the drone 10 has a predetermined response characteristic, the operator determines that the drone 10 is not malfunctioning, and inputs data indicating that the drone 10 is not malfunctioning via the input device 228. input into computer 210; The computer 210 corresponds data indicating that the drone 10 is not malfunctioning and a plurality of combinations (for example, the above N combinations) of the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10. The data is then stored in the secondary storage device 222 as teacher data. In this case, N pieces of teacher data are stored in the secondary storage device 222. The ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 are input data of teacher data, and are an example of information for specifying data related to a failure. The data indicating that the drone 10 is not malfunctioning is output data of the teacher data, and is an example of data regarding the malfunction.

On the other hand, if the operator determines that the drone 10 does not have the predetermined response characteristics, the operator determines that the drone 10 is malfunctioning, inspects the drone 10, and discovers the malfunction location. Then, the operator inputs data indicating that the drone 10 is malfunctioning, data indicating the location of the malfunction, and data indicating measures to be taken against the malfunction into the computer 210 via the input device 228. The computer 210 stores data indicating that the drone 10 is malfunctioning, data indicating the location of the malfunction, data indicating countermeasures against the malfunction, an ideal position and attitude of the drone 10, and an actual position and attitude of the drone 10. A plurality of combinations (for example, the above N combinations) are stored in the secondary storage device 222 as teacher data. In this case, N pieces of teacher data are stored in the secondary storage device 222. The ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 are input data of teacher data, and are an example of information for specifying data related to a failure. The data indicating that the drone 10 is malfunctioning, the data indicating the location of the malfunction, and the data indicating countermeasures against the malfunction are output data of the teacher data and are examples of data regarding the malfunction.

Let's further explain the training data. The teacher data includes input data and output data, as described above.

The input data of the teacher data includes an ideal movement state of the drone 10 and an actual movement state of the drone 10 according to a signal indicating the content of movement from the operating device 500. As described above, the movement state is the position and attitude of the drone 10, and the attitude is represented by the yaw angle, pitch angle, and roll angle of the drone 10. Therefore, the input data of the teaching data includes data on the ideal and actual position of the drone 10, yaw angle, pitch angle, and roll angle of the drone 10.

More specifically, the input data includes, for example, the ideal three-dimensional space position (X0, Y0, Z0) of the center of the main body 12 of the drone 10 and the actual position of the drone 10, respectively. The input includes the position (Xa, Ya, Za) in the three-dimensional space. In addition, the input data includes yaw angles (YAW0, YAWa), which are rotation angles around each axis, in three-dimensional coordinates based on the center of the main body 12 of the drone 10, and pitch angles ( P0, Pa) and roll angle (R0, Ra).

Note that the yaw angle is, for example, an angle in which clockwise rotation is positive based on the traveling direction of the drone 10 when the drone 10 is viewed from above. The pitch angle is, for example, an angle in which the counterclockwise rotation is positive with respect to the horizontal when the drone 10 is viewed from the left side. The roll angle is, for example, an angle in which the counterclockwise rotation is positive with respect to the horizontal when the drone 10 is viewed from the direction of travel.

The output data of the teacher data includes data regarding the failure of the drone 10. The data regarding the malfunction of the drone 10 includes data indicating whether or not the drone 10 is malfunctioning. Furthermore, if the drone 10 is out of order, the data regarding the failure includes data on the location of the failure and data indicating countermeasures against the failure.

　Explain more specific teacher data. As an example, the following first teaching data to seventh teaching data will be explained.

(First teacher data)
The first teaching data will be explained.
A limited flight test was conducted in which the aircraft ascended directly upward from the position (X, Y, Z) = (0, 0, 0) in a horizontal state at a climbing speed of Vh. However, in reality, the drone 10 rose horizontally, but the rising speed was Vh/5.

The input data of the first teacher data includes the ideal position and attitude of the drone 10, "X0=0, Y0=0, Z0=0, rising speed=Vh, YAW0=0, P0=0, R0= 0'' and the actual position and attitude of the drone 10, ``Xa=0, Y0=a, Za=0, rising speed=Vh/5, YAWa=0, Pa=0, Ra=0''. be.

In this way, there was an instruction to "climb directly overhead in a horizontal state at a climbing speed of Vh," but in reality, the drone 10 rose horizontally but at a climbing speed of Vh/5. The propeller motors were not at fault because the aircraft climbed horizontally, but the formula for calculating the current to be output to each motor depending on the content of the movement, which was set in the flight controller, was incorrect.

Therefore, the output data of the first teacher data is as follows. That is, the output data includes data indicating that the drone 10 is malfunctioning, data indicating the "flight controller" as data on the location of the malfunction, and data indicating "adjusting the calculation formula of the flight controller" as data indicating the countermeasure. It is data.

In this way, it is understood that there is a correlation between the input data of the first teacher data and the output data of the first teacher data.

(Second teacher data)
The second training data will be explained.
A limited flight test was conducted in which the aircraft climbed straight upward from the position (X, Y, Z) = (0, 0, 0) in a horizontal state at a climbing speed Vh, and then tilted counterclockwise at a roll angle ry. However, in reality, although the vehicle ascended at approximately the ascending speed Vh, when ascending, the vehicle tilted clockwise at a roll angle rx, and when tilting counterclockwise, the vehicle tilted at a roll angle ry−rx.

The input data of the second teacher data includes the ideal position and attitude of the drone 10, "X0=0, Y0=0, Z0=0, rising speed=Vh, YAW0=0, P0=0, R0= 0" → "R0=ry" and the actual position and attitude of the drone 10, "X0=0, Y0=0, Z0=0, rising speed=Vh, YAW0=0, P0=0, R0=- rx” → “R0=ry-rx”.

In this way, there was an instruction to "climb straight upwards in a horizontal state at a climbing speed Vh, and then tilt counterclockwise at a roll angle rx," but in reality, when climbing, the roll angle is clockwise. The reason why the rx tilted and the roll angle ry-rx tilted counterclockwise was because the calibration was inaccurate. Specifically, during calibration, the roll angle was set clockwise. This is because the sensor that detects the tilt angle has been calibrated so that it is horizontal even though it is tilted at ry.
Note that calibration refers to measuring the bias of the sensor using a standard device or the like and adjusting it to obtain the correct value in order to obtain a standard value with the sensor.

Therefore, the output data of the second teacher data is as follows. That is, the output data includes data indicating that the drone 10 is malfunctioning, data indicating the "sensor that detects the tilt angle" as data on the malfunction location, and data indicating the "sensor that detects the tilt angle" as data indicating the countermeasure. This data shows that the system is being reworked.

In this way, it is understood that there is a correlation between the input data of the second teacher data and the output data of the second teacher data.

(Third teacher data)
The third training data will be explained.
A limited flight test was conducted in which the aircraft stopped in the air at position (X, Y, Z) = (0, 0, H). However, in reality, the drone 10 had Xa=0, Ya=0, Za=HH, YAWa=0, Pa=0, and Ra=0. This means that although the drone 10 actually had the ideal attitude, the height (Z) was insufficient by h.

The input data of the third training data includes "X0=0, Y0=0, Z0=H, YAW0=0, P0=0, R0=0", which is the ideal position and attitude of the drone 10, and There are ten actual positions and orientations: "Xa=0, Ya=0, Za=Hh, YAWa=0, Pa=0, Ra=0".

In this way, the flight was supposed to be stopped in the air at the position (X, Y, Z) = (0, 0, H), but there was no problem with the attitude, but the flight was short in height by 4. This is because the rotation speeds of the two propellers are insufficient. The reason why the rotation speeds of the four propellers were insufficient evenly was due to a failure in the power supply that caused an insufficient output.

Therefore, the output data of the third teacher data is as follows. That is, the output data includes data indicating that the drone 10 is out of order, data indicating "power source" as data on the location of the failure, and data indicating "replacement of power source" as data indicating countermeasures.

In this way, it is understood that there is a correlation between the input data of the third teacher data and the output data of the third teacher data.

(Fourth teacher data)
The fourth training data will be explained.
A limited flight test was conducted in which the aircraft stopped in the air at position (X, Y, Z) = (0, 0, H). However, in reality, the drone 10 had Xa=0, Ya=0, Za=H+h, YAWa=0, Pa=0, and Ra=0. This means that although the attitude of the drone 10 was actually ideal, the height (Z) was h too high.

The input data of the fourth teacher data includes "X0=0, Y0=0, Z0=H, YAW0=0, P0=0, R0=0", which is the ideal position and attitude of the drone 10, and There are ten actual positions and orientations, "Xa=0, Ya=0, Za=H+h, YAWa=0, Pa=0, Ra=0."

In this way, the flight was supposed to be suspended in the air at the position (X, Y, Z) = (0, 0, H), but there was no problem with the attitude, but the height was h too high. This is because the rotational speeds of the four propellers were equally excessive. The reason why the rotational speeds of the four propellers were uniformly excessive was due to a failure in the power supply causing excessive output.

Therefore, the output data of the fourth teacher data is as follows. That is, data indicating that the drone 10 is out of order, data indicating "power supply" as data on the location of the failure, and data indicating "replacement of power supply" as data indicating countermeasures.

In this way, it is understood that there is a correlation between the input data of the fourth teacher data and the output data of the fourth teacher data.

(Fifth teacher data)
The fifth training data will be explained.
A limited flight test was conducted in which the aircraft flew counterclockwise at a height H at a position (X, Y, Z) = (0, 0, H) and a counterclockwise roll angle r0. In addition, in the drone 10 of the present embodiment, when flying in which the counterclockwise roll angle increases, the rotation speed of the right front propeller and the rotation speed of the right rear propeller are changed to the rotation speed of the left front propeller and the rotation speed of the left rear propeller. This is done by making the rotation speed smaller than that of the side propeller.
However, in reality, the drone 10 had Xa=0, Ya=0, Za=H+h0, YAWa=0, Pa=0, and Ra=r0-r. This means that the height was high by h0 and the counterclockwise roll angle was short of the r angle.

The input data of the fifth teacher data includes "X0=0, Y0=0, Z0=H, YAW0=0, P0=0, R0=r0", which is the ideal position and attitude of the drone 10, and There are ten actual positions and orientations: "Xa=0, Ya=0, Za=H+h0, YAWa=0, Pa=0, Ra=r0-r".

The reason for this kind of flight where the height was high h0 and the counterclockwise roll angle was insufficient r angle was due to the excessive rotation speed of the right front propeller and the excessive rotation speed of the right rear propeller. This is because there was The reason why there was an excessive rotation speed of the right front propeller and an excessive rotation speed of the right rear propeller was due to a failure in which the rotation speed of the right front propeller motor and the right rear propeller motor was insufficient. This is because.

Therefore, the output data of the fifth teacher data is as follows. In other words, data indicating that the drone 10 is malfunctioning, data indicating the failure location as "the right front propeller motor and right rear propeller motor", and data indicating the countermeasure as "the right front propeller motor". This data shows the replacement of the motor and the right rear propeller motor.

In this way, it is understood that there is a correlation between the input data of the fifth teacher data and the output data of the fifth teacher data.

(Sixth teacher data)
The sixth training data will be explained.
A limited flight test was conducted in which the front side of the drone 10 was tilted downward at an angle of p0 at the position (X, Y, Z) = (0, 0, H). In addition, in the drone 10 of this embodiment, when the front side of the drone 10 is tilted downward, the rotation speed of the right front propeller and the rotation speed of the left front propeller are changed to the rotation speed of the right rear propeller and the left rear propeller. This is done by making the rotation speed smaller than that of the side propeller.
However, in reality, the drone 10 has Xa=0, Ya=0, Za=H+h0, YAWa=0, Pa=p0-p, and Ra=0. This means that the height was higher than h0, and the pitch angle at which the front side of the drone 10 tilted downward was insufficient to the p angle.

The input data of the sixth teacher data includes "X0=0, Y0=0, Z0=H, YAW0=0, P0=p0, R0=0", which is the ideal position and attitude of the drone 10, and There are ten actual positions and orientations: "Xa=0, Ya=0, Za=H+h0, YAWa=0, Pa=p0-p, Ra=0".

The reason why the height was high h0 and the pitch angle where the front side of the drone 10 tilted downward was insufficient for the p angle was because the rotational speed of the front right propeller was too high and the front left propeller was too fast. This is because the rotational speed was excessive. The reason why the rotation speed of the right front propeller became excessive and the rotation speed of the left front propeller became excessive is that a failure occurred in the right front propeller motor and the left front propeller motor causing the rotation speed to be excessive. This is because.

Therefore, the output data of the sixth teacher data is as follows. In other words, data indicating that the drone 10 is malfunctioning, data indicating "the right front propeller motor and left front propeller motor" as data on the malfunction location, and data indicating "the right front propeller motor" as data indicating the countermeasure. This data indicates "replacement of the motor and front left propeller motor."

In this way, it is understood that there is a correlation between the input data of the sixth teacher data and the output data of the sixth teacher data.

(7th teacher data)
The seventh training data will be explained.
A limited flight test was conducted in which the drone 10 flew at the position (X, Y, Z) = (0, 0, H) with the direction of the drone 10 facing to the right at the y0 angle. In addition, in the drone 10 of this embodiment, when the drone 10 flies in the right direction, the rotation speed of the right rear propeller and the rotation speed of the left front propeller are changed to the rotation speed of the right front propeller and the left propeller. This is done by increasing the rotational speed of the rear propeller.
However, in reality, the drone 10 has Xa=0, Ya=0, Za=H−h0, YAWa=y0−y, Pa0=, and Ra=0. This means that the height was short by h0, and the rightward angle of the drone 10 was short by y angle.

The input data of the seventh teacher data includes "X0=0, Y0=0, Z0=H, YAW0=y0, P0=0, R0=0", which is the ideal position and attitude of the drone 10, and There are ten actual positions and orientations, "Xa=0, Ya=0, Za=H−h0, YAWa=y0−y, Pa0=, Ra=0."

The reason for this flight was that the height was insufficient h0 and the rightward angle of the drone 10 was insufficient y angle, because the rotation speed of the right rear propeller was insufficient and the rotation speed of the left front propeller was insufficient. This is because the speed was insufficient. The reason why the rotation speed of the right rear propeller was insufficient and the rotation speed of the left front propeller was insufficient was that a failure occurred in the right rear propeller motor and the left front propeller motor that caused the rotation speed to be insufficient. It is.

Therefore, the output data of the seventh teacher data is as follows. In other words, data indicating that the drone 10 is malfunctioning, data indicating the failure location as "right rear propeller motor and left front propeller motor", and data indicating countermeasures as "right rear propeller motor". This data indicates the replacement of the propeller motor and the front left propeller motor.

In this way, it is understood that there is a correlation between the input data of the seventh teacher data and the output data of the seventh teacher data.

During the limited flight test for the third to seventh teacher data, it is assumed that the flight controller is normal and that each sensor is calibrated accurately.

The first teacher data to the seventh teacher data are examples, and the teacher data is not limited to the first teacher data to the seventh teacher data. Get data.

In step 500SA, the acquisition unit 500A in the learning unit 500 specifically inputs the information for specifying the data related to the failure stored in the secondary storage device 222 as described above, and Obtain multiple pieces of training data whose output data is data related to . As described above, the information for specifying the data regarding the failure includes the ideal movement state (position and attitude) of the drone 10 according to the signal indicating the movement details from the operating device 500 and the actual movement state of the drone 10. This is data indicating the movement state (position and orientation).

In step 500SB, the learning processing unit 500B of the learning unit 500 uses the teacher data to learn a model 222M that inputs information for specifying data regarding the failure of the drone 10 and outputs data regarding the failure of the drone 10. do. The learning processing unit 500B inputs information for specifying data regarding the failure of the drone 10 in the teaching data to the input layer, and inputs information to the intermediate layer so that the data regarding the failure in the drone 10 in the teaching data is output from the output layer. learn the parameters of

As explained above, the learning unit 500 generates the learned model 222M by learning the model 222M using the teacher data. By the way, the models 222M that the learning unit 500 learns include, firstly, a model that has not been trained at all, and secondly, a trained model 222M. Therefore, by learning a model that has not been trained at all by the learning unit 500, a trained model 222M that has been trained for the first time is generated. Furthermore, by the learning unit 500 learning the learned model 222M, a further learned model 222M is generated.

In the example described above, the ideal movement state (position and attitude) of the drone 10 and the actual movement state (position and attitude) of the drone 10 are used as information for identifying data related to a failure. The technology is not limited to this. For example, instead of the ideal movement state of the drone 10 and the actual movement state of the drone 10, a value indicating the degree of coincidence between the ideal movement state of the drone 10 and the actual movement state of the drone 10 may be used.

Here, the value indicating the degree of coincidence will be explained. Values indicating the degree of coincidence include first to fifth values.
The first to fifth values indicating the degree of coincidence are an example of "information for specifying data related to a failure" of the technology of the present disclosure.

The first value indicating the degree of coincidence is a value indicating the degree of coincidence between the ideal three-dimensional space position of the drone 10 and the actual three-dimensional space position after a predetermined period of time has passed since receiving the instruction signal. It is.

The center of the mounting table 114 of the original aircraft holding device 150 when the support column 110 is not extended is the origin of the three-dimensional space, specifically, the center of the three directions (X direction, Y direction, and Z direction) in the three-dimensional space. (i.e., the origin). In a plane passing through the center of the platform 114 of the aircraft holding device 150, a predetermined direction passing through the center of the platform 114 is defined as the X direction, and a direction perpendicular to the X direction is defined as the Y direction. The Z direction is defined as a direction (vertical direction, that is, height direction) perpendicular to each of the X direction and the Y direction.

The first value indicating the degree of coincidence is the difference between the ideal three-dimensional space position (X0, Y0, Z0) of the center of the main body 12 of the drone 10 when a predetermined period of time has passed since receiving the instruction signal and the actual three-dimensional position (X0, Y0, Z0). This is the degree of coincidence with the position (Xa, Ya, Za) in the dimensional space. To explain in more detail, the first value is Xa/X0, Ya/Y0, Za/Z0, and their average value when a predetermined time has elapsed since receiving the instruction signal.

The second to fourth values indicating the degree of coincidence are values indicating the degree of coincidence between the ideal posture and the actual posture of the drone 10 when a predetermined period of time has passed since the instruction signal was received. The attitude is represented by a yaw angle, a pitch angle, and a roll angle, which are rotation angles around each axis, in three-dimensional coordinates based on the center of the main body 12 of the drone 10. If the ideal yaw angle of the drone 10 when a predetermined time has elapsed after receiving the instruction signal is Y0, and the actual yaw angle is Ya, then the second value is Ya/Y0. Assuming that the ideal pitch angle of the drone 10 when a predetermined time has elapsed after receiving the instruction signal is P0, and the actual pitch angle is Pa, the third value is Pa/P0. Assuming that the ideal roll angle of the drone 10 after a predetermined period of time has passed since receiving the instruction signal is R0, and the actual roll angle is Ra, the fourth value is Ra/R0.

The fifth value indicating the degree of coincidence is a value indicating the degree of coincidence between the image of the actual state of the drone 10 shown in the area 224S2 and the image of the ideal state of the drone 10 shown in the area 224S1. Specifically, a plurality of corresponding points of the drone 10 are extracted from each of the image of the actual state of the drone 10 and the image of the ideal state of the drone 10. Examples of the plurality of corresponding points on the drone 10 include the center point of each propeller and the center point of the main body. Next, for example, a plurality of points extracted from the image of the drone 10 in its actual state are superimposed on the image of the drone 10 in its ideal state. The statistical value between the plurality of points extracted from the image of the actual state of the drone 10 and superimposed on the image of the ideal state and the plurality of points extracted from the image of the ideal state is determined by a fifth index indicating the degree of coincidence. Calculate as the value of .

Specifically, the centers of the first to fourth propellers extracted from the image of the actual state of the drone 10 are superimposed on the image of the ideal state of the drone 10, and the centers of the superimposed first to fourth propellers are Statistical values are calculated between the points and the center points of the first to fourth propellers extracted from the image of the ideal state of the drone 10. Specifically, the statistical value is the distance between each of the center points of the superimposed first to fourth propellers and each of the center points of the first to fourth propellers extracted from the ideal state image. These include the total, the maximum value of the distance, and the value obtained by subtracting the minimum value from the maximum value of the distance.

Note that the fifth value may be calculated by superimposing a plurality of points extracted from the image of the ideal state of the drone 10 on the image of the actual state of the drone 10.

(Identification of data regarding failure of drone 10)
Next, a program for processing to specify data regarding a failure of the drone 10 will be explained. The process of identifying the data regarding the failure of the drone 10 includes, firstly, a first process of identifying the data regarding the failure of the drone 10 when instructed by the operator who has viewed the information for identifying the data regarding the failure. (See FIG. 8), and secondly, a second process (see FIG. 12) that automatically identifies data related to a failure of the drone 10.

As mentioned above, you can see the display of the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 (Figures 9 to 11), and see the flight status of the drone 10 in the above limited flight test. Then, the operator can determine whether the drone 10 has a predetermined response characteristic. If the operator determines that the drone 10 does not have a predetermined response characteristic, the operator may determine that the drone 10 is malfunctioning, and may inspect the drone 10 to discover the malfunction location.

However, in the present embodiment, the information output device 170 executes a program for processing to specify data related to a failure of the drone 10.

FIG. 8 shows a flowchart of the identification program 222P3 of the first process for identifying data related to a failure, which is executed by the CPU 212 of the information output device 170. The identification program 222P3 of the first process for identifying data related to a failure shown in FIG. 8 starts when an execution instruction button (not shown) is turned on.

In step 111, the reading unit 512 reads information for specifying the data regarding the failure of the drone 10 from the secondary storage device 222, and the display unit 514 displays the read information for specifying the data regarding the failure of the drone 10. is displayed on the display 224 as shown in FIGS. 9 to 11.

FIG. 9 shows a screen 224S of the display 224 that displays information for specifying data regarding a malfunction of the drone 10. As shown in FIG. 9, the screen 224 has three areas 224S1, 224S2, and 224S3.

Area 224S1 will be explained. The processing unit 502 converts each image of the drone 10 taken in advance by each of the cameras 202N1 to 202N4 into correspondence with the ideal movement state (position and attitude) of the drone 10 when a predetermined period of time has elapsed since receiving the instruction signal. Transform as you do. The processing unit 502 switches each of the cameras 202N1 to 202N4 to the area 224S1 and displays the transformed image (that is, the image of the ideal movement state (that is, the position and attitude) of the drone 10) after receiving the instruction signal. The corresponding display is displayed when a predetermined period of time has elapsed since then.

Next, the area 224S2 will be explained. The processing unit 502 sends images of the actual movement state (i.e., position and orientation) of the drone 10 obtained by the cameras 202N1 to 202N4 after a predetermined period of time has passed since receiving the instruction signal to the area 224S2. - Switch and display every 202N4.

The more the image of the actual state of the drone 10 shown in the area 224S2 differs in position and attitude from the image of the ideal state of the drone 10 shown in the area 224S1, the more the actual response characteristics of the drone 10 differ from the image of the ideal state of the drone 10 shown in the area 224S1. It is understood that this is worse than the 10 types of above-mentioned stored response characteristics.

In the above example, the processing unit 502 switches between the area 224S1 and the area 224S2 for each of the cameras 202N1 to 202N4 to display the transformed image and the actual image of the drone 10, but the technology of the present disclosure is not limited to this. For example, each region of area 224S1 and area 224S2 is divided by the number of cameras so as to correspond to cameras 202N1 to 202N4, and each of the transformed images and each actual image of drone 10 are displayed in each region. It's okay.

The image of the actual state of the drone 10 shown in the area 224S2 and the image of the ideal state of the drone 10 shown in the area 224S1 are examples of "information for identifying data related to failure" of the technology of the present disclosure.

The area 224S3 is an area that displays a value indicating the degree of agreement between the ideal position and attitude of the drone 10 and the actual position and attitude when a predetermined time has elapsed since receiving the instruction signal.

The area 224S3 may display the first to fifth values indicating the degree of matching, as well as an explanation of the values. For example, the first value is described as "a value indicating the degree of agreement between the ideal three-dimensional space position of the drone 10 and the actual three-dimensional space position." Note that the explanation of the second value to the fifth value is also the definition of each value, similar to the explanation of the first value.

It is understood that the more each of the first to fourth values displayed in the area 224S3 differs from 1, the worse the actual response characteristics of the drone 10 are than the stored response characteristics of the type of the drone 10. be done. It is understood that the larger the fifth value displayed in the area 224S3, the worse the actual response characteristics of the drone 10 are than the stored response characteristics of the type of the drone 10.

FIG. 10 shows the height of the drone 10 relative to the elapsed time from when the start button is turned on to when the stop button is turned on since the information output processing program 222P1 is started, which is displayed on the display 224. A graph of changes in is shown. A time change in the ideal height of the drone 10 is shown by a solid line, and a time change in the actual height of the drone 10 is shown by a dotted line. It is understood that the more the time change of the actual height of the drone 10 shown by the dotted line differs from the time change of the ideal height of the drone 10 shown by the solid line, the worse the response characteristics of the drone 10 are. . Specifically, in the example shown in FIG. 9, the content of the movement indicated by the instruction signal transmitted from the operating device 50 is to gradually raise the support 110 from the initial time when it is not extended, as shown by the solid line. The device flies in a stopped position (that is, hovers) at a predetermined height. However, in the actual movement of the drone 10, as shown by the dotted line, from the initial time when the support 110 is not extended, the drone 10 gradually rises and flies to a halt in the air, but the rising speed and the height at which it flies to a halt in the air are fixed at a predetermined level. It was lower than the height. Therefore, it is understood that the actual response characteristic of the climbing speed of the drone 10 is worse than the response characteristic of the above-mentioned stored climbing speed of the type of drone 10.

FIG. 11 shows the pitch angle of the drone 10 relative to the elapsed time from when the start button is turned on to when the stop button is turned on since the information output processing program 222P1 is started, which is displayed on the display 224. A graph of changes in is shown. A time change in the ideal pitch angle of the drone 10 is shown by a solid line, and a time change in the actual pitch angle of the drone 10 is shown by a dotted line. It is understood that the more the time change of the actual pitch angle of the drone 10 shown by the dotted line differs from the time change of the ideal pitch angle of the drone 10 shown by the solid line, the worse the response characteristics of the drone 10 are. . Specifically, in the example shown in FIG. 10, the content of the movement indicated by the instruction signal transmitted from the operating device 50 is such that the pitch angle gradually changes from when the start button is turned on, as shown by the solid line. It becomes large and remains almost unchanged at a given angle. However, in the actual movement of the drone 10, as shown by the dotted line, the pitch angle gradually increases from when the start button is turned on, but the increasing speed is the content of the movement indicated by the instruction signal. The timing at which the angle becomes smaller and becomes almost unchanged at a predetermined angle is delayed from the content of the movement indicated by the instruction signal. Therefore, it is understood that the actual response characteristic of the pitch angle of the drone 10 is worse than the response characteristic of the above-mentioned stored climbing speed of the type of the drone 10.
Note that other roll angles and yaw angles may also be displayed in the same manner as the pitch angle.

In step 115, the processing unit 502 determines whether or not it has been instructed to identify the cause.

As mentioned above, the operator selects images of the ideal and actual movement states of the drone, values indicating the degree of agreement between the ideal position and attitude of the drone 10 and the actual position and attitude (see FIG. 9), and the drone. Check the graph of the change in the height of the drone 10 (see Figure 10) and the graph of the change in the pitch angle of the drone 10 (see Figure 11), and check the actual response characteristics of the drone 10 by checking the graph of the change in the height of the drone 10 (see Figure 10), and determine the actual response characteristics of the drone 10. It can be evaluated in relation to the response characteristics of the rising speed. Then, when the operator evaluates the actual response characteristics of the drone 10 and evaluates that the actual response characteristics of the drone 10 are worse than the response characteristics of the above-mentioned stored climbing speed of the type of the drone 10, the operator determines the cause of the bad response. Decide whether to identify. If it is determined to identify the cause of poor actual response characteristics of the drone 10, a specific execution button (not shown) is turned on.

If the specific execution button is turned on in this manner, the determination in step 115 is affirmative, and the first process proceeds to step 117. If the specific execution button is not turned on, a negative determination is made in step 115, and the first process ends.

In step 117, the identifying unit 516 identifies the cause of the failure and a countermeasure using the learned model 222M. That is, the specifying unit 516 specifies data related to the failure of the drone 10 based on the ideal moving state of the drone 10 and the actual moving state of the drone 10 according to the signal, and the learned model 222M. More specifically, the specifying unit 516 inputs the ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the above-mentioned signals to the input layer of the model 222M, and inputs the ideal movement state of the drone 10 and the actual movement state of the drone 10 from the output layer of the model 222M. Output data related to failures. The ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the above signals input to the input layer of the model 222M are N combinations of the ideal position and attitude and the actual position and attitude of the drone 10. It is one of the The N combinations are the N combinations that were most recently stored in the secondary storage device 222 by executing the information processing program 222P1 in FIG. 7A before the specific program 222P3 was executed this time.

At step 119, the display unit 514 displays the cause of the failure and countermeasures on the display 224.

Next, with reference to FIG. 12, a second process for automatically identifying data related to a failure of the drone 10 will be described. FIG. 12 shows a flowchart of a second processing identification program that is executed by the CPU 212 of the information output device 170 to automatically identify data related to a failure of the drone 10.

In step 121, the reading unit 512 reads, from the secondary storage device 222, a value indicating the degree of matching among the information for specifying the data related to the failure of the drone 10, and in step 123, the processing unit 502, It is determined whether the read value indicating the degree of matching is outside a predetermined allowable range.

If the value indicating the degree of coincidence is outside the predetermined allowable range, the determination in step 123 is affirmative, and the second process proceeds to step 117. On the other hand, if the value indicating the degree of coincidence is within a predetermined allowable range, the determination in step 123 is negative, and the second process ends.

The processing in step 117 and step 119 is the same as step 117 and step 119 in the first processing, so the explanation thereof will be omitted.

As described above, in this embodiment, it is possible to specify data related to failures of the drone 10 manufactured according to a completed design so as to have predetermined response characteristics.

In the present embodiment, data regarding the failure of the drone 10 is specified using a plurality of combinations of information for identifying data regarding the failure of the drone 10 stored in the secondary storage device 22 and data regarding the failure of the drone 10. Compared to this, it is possible to identify data related to aircraft failures even in patterns that do not exist in combinations.

In the present embodiment, it is possible to provide a learning method, a learning device, and a program that can generate a learned model 222M for specifying data regarding a failure of the drone 10.

In this embodiment, a model for specifying data regarding a failure of the drone 10 can be provided.

As described above, the aircraft response characteristic providing system of the present embodiment is capable of controlling the flight of the drone 10 when the operator operates the operating device 50 to fly the drone 10 while holding it on the mounting base 114. can provide, in particular store, information to identify data regarding the failure of the device. More specifically, in the present embodiment, firstly, the movement state of the drone 10 and the detected movement state of the drone 10 are determined based on the content of the movement instructed by the operating device 50; A value indicating the degree of coincidence between the movement state of the drone 10 based on the contents of the instructed movement and the detected movement state of the drone 10 can be stored. In particular, in this embodiment, the position and attitude of the drone 10 can be stored as the movement state, and furthermore, the yaw angle, pitch angle, and roll angle of the drone 10 can be stored as the attitude.
The present embodiment can provide an aircraft holding device used in the aircraft response characteristics providing system and an information output device used in the aircraft response characteristics providing system.

Further, the aircraft holding device 150 of the present embodiment can hold the drone 10 on the mounting table 114 so as to be rotatable three-dimensionally. Therefore, information for identifying data related to flight failures of the drone 10 can be stored without actually performing a test flight. Therefore, even if the assembled drone 10 does not have response characteristics appropriate for the type of drone 10, it may crash and be destroyed during a movement test (i.e., flight test). This can be prevented.

In this embodiment, the support of the aircraft holding device 150 is expandable and retractable, so the drone 10 can move in the vertical direction. In this embodiment, one end of the support column 110 and the mounting table 114 are connected by a universal joint, so that they can be rotated three-dimensionally. Therefore, the present embodiment can provide information for specifying data regarding failures in vertical movement and three-dimensional rotation of the drone 10.

In this embodiment, the lower surface of the base 118 of the aircraft holding device 150 is provided with a plurality of casters 122N1 to 122N4, so that the aircraft holding device 150 can move as the drone 10 moves horizontally or diagonally upward or downward. , along the plane where the aircraft holding device 150 is provided. Therefore, the present embodiment can provide information for specifying data regarding a failure in horizontal movement or diagonally upward or downward movement of the drone 10.

In this embodiment, a learned model is used to identify the cause of a failure and a countermeasure, but the technology of the present disclosure is not limited to this. For example, a plurality of combinations of an ideal moving state of the drone 10, an actual moving state of the drone 10, causes of failure of the drone 10, and countermeasures are stored in the secondary storage device 222. The identification unit 516 determines whether the drone 10 is malfunctioning based on the plurality of combinations stored in the secondary storage device 222 and the ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the above-mentioned signals. Identify the cause and countermeasures. More specifically, the specifying unit 516 selects a combination that corresponds to the ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the signal from among the plurality of combinations stored in the secondary storage device 222. Identify the cause of the failure of the drone 10 and a countermeasure plan. The display unit 514 displays the identified cause of the failure of the drone 10 and countermeasures on the display 224. In this way, data regarding the failure of the drone 10 can be specified. In particular, in this example, model learning can be made unnecessary.

In the embodiment described above, in step 620, the storage unit 510 stores the ideal position and orientation of the drone 10 and the actual position and orientation of the drone 10 in the secondary storage device 222 together with each image. By outputting it, it is stored in the secondary storage device 222. The technology of the present disclosure is not limited to this. For example, in step 620, instead of outputting the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 to the secondary storage device 222 together with each image, or 222 , the display unit 514 may also output to the display 224 . Thereby, the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 can be displayed in real time on the display 224 in correspondence with each other. In this manner, the technology of the present disclosure stores the ideal position and attitude of the drone 10 and the actual position and attitude of the drone 10 in correspondence with each other in the secondary storage device 222 or displays them on the display 224. In other words, it is possible to provide information for identifying data related to flight failures of the drone 10.

The secondary storage device 222 and the display 224 are examples of the "output unit" of the technology of the present disclosure.

In the embodiment described above, in

steps

610 and 618, the calculation unit 508 calculates the position and orientation of the drone 10 from each image captured from each of the cameras 202N1 to 202N4 of the information output device 170. The technology of the present disclosure is not limited to this. For example, in the drone 10, the actual position of the drone 10 may be detected using GPS (Global Positioning System), or the angles may be detected using various sensors that detect the yaw angle, pitch angle, and roll angle. You can also

In this way, when detecting the actual position of the drone 10 using GPS or detecting these angles using various sensors that detect the yaw angle, pitch angle, and roll angle, the cameras 202N1 to 202N4 Also, the processes of steps 604 to 608 and 616 may be omitted. However, without omitting the cameras 202N1 to 202N4, the position and attitude of the drone 10 are calculated from each image captured from each of the cameras 202N1 to 202N4, and the position and orientation of the drone 10 is calculated using the above-mentioned GPS and various sensors. and posture, and calculate the average value of each, and use the average value.

In this embodiment and the example, the actual movement state of the aircraft detected by the detection unit (cameras 202N1 to 202N4, various sensors) provided in at least one of the drone 10 and the information output device 170 can be used. .

The actual direction is an example of the "ideal movement state of the aircraft in response to the instruction signal" of the technology of the present disclosure, and the ideal direction is an example of the "actual movement state of the aircraft" of the technology of the present disclosure. .

(Second embodiment)
Next, a second embodiment will be described. The configuration of the second embodiment is substantially the same as the configuration of the first embodiment, so the same parts are given the same reference numerals, the explanation thereof will be omitted, and the different parts will be explained.

The drone 10 of the first embodiment is manufactured according to a completed design so as to have predetermined response characteristics. In contrast, the drone of the second embodiment is a drone that is currently being designed to have predetermined response characteristics.

The model 222M stored in the secondary storage device 222 of the first embodiment is a model for specifying data regarding a failure of the drone 10. On the other hand, the model 222M stored in the secondary storage device 222 of the second embodiment is a model for specifying data related to improvement of the drone at the stage of designing it to have predetermined response characteristics. It's a model.
Data regarding failures and data regarding improvements to the drone are examples of "data regarding performance" in this embodiment.

The operation of the second embodiment will be explained.

(Limited flight test)
First, the information output device 170 is designed to have a predetermined response characteristic by conducting a limited flight test on a drone that is currently being designed to have a predetermined response characteristic. Obtain information to identify data on drone improvements in stages. Therefore, the information output processing of the information output processing program 222P1 shown in FIG. 7A is performed on the drone that is currently being designed to have predetermined response characteristics. Note that the information output processing of the information output processing program 222P1 shown in FIG. 7A of the second embodiment is the same as that of the first embodiment described above, so a description thereof will be omitted.

(Model learning)
Second, the information output device 170 learns the model using information for specifying data related to improvement of the drone 10, which is obtained by performing such a limited flight test. Therefore, the learning processing program 222P2 shown in FIG. 7B was obtained by conducting a limited flight test conducted on a drone at the stage of designing the learning processing program 222P2 to have predetermined response characteristics. Train a model using data. Note that the learning process of the learning process program 222P2 shown in FIG. 7B of the second embodiment is the same as that of the first embodiment described above, so a description thereof will be omitted.

(Identification of data regarding failure of drone 10)
Thirdly, the information output device 170 uses the learned model to identify data regarding improvements to the drone at the stage of designing it to have predetermined response characteristics. Therefore, by executing the specific program for the first process shown in FIG. 8 or the specific program for the second process shown in FIG. Identify data. Note that the specifying process of the specifying program (FIG. 8 or 12) in the second embodiment is the same as that in the first embodiment described above, so a description thereof will be omitted. Note that the identification process of the second embodiment is executed using the "failure-related data" of the identification program (FIG. 8 or 12) of the first embodiment as "improvement-related data."

(teaching data)
Next, the teaching data will be explained. As described above, the model learning in the second embodiment is performed at the stage where the structure of the drone is designed to have predetermined response characteristics.

(Teaching data A)
The predetermined response characteristic is, for example, that the time TR until the rightward roll angle reaches r1 is Tr100 (seconds).
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC1, the slope of the relational expression of the rotational speed of the motor shaft with respect to the supplied current is determined so that the response characteristics are as described above. A drone was provisionally designed to use motor 3 of A=A3, and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr99 (<Tr100).
For motor 3, in a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC1, the time TR until the roll angle to the right reaches r1 is within Tr100. It was found that it has the performance that can be achieved. Therefore, the data regarding improvement is "no improvement required".

From the above, the input data of the teacher data A is the propeller length LLx = LL1, the distance between the propeller rotation center and the body's center of gravity position LCx = LC1, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 2. The slope A=A2.
The output data of the teacher data A is "no improvement required".

(Teaching data B)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC1, the slope of the relational expression of the rotational speed of the motor shaft with respect to the supplied current is determined so that the response characteristics are as described above. A drone was provisionally designed to use a motor 2 with A=A2 (<A3), and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr110 (>Tr100).
The motor 2 has a structure in which the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC1, and the performance is such that the time TR until the roll angle to the right reaches r1 is within Tr100. It was found that it does not have
As a result of replacing motor 2 with motor 3 and conducting a limited flight test again, the time taken for the roll angle to the right to reach r1 was TR=Tr099 (<Tr100). Therefore, the data regarding improvement are "improvement required" and "replace motor 2 with motor 3."

From the above, the input data for teacher data B is the propeller length LLx = LL1, the distance between the propeller rotation center and the body's center of gravity position LCx = LC1, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 2. The slope A=A2.
The output data of teacher data B are "Improvement required" and "Replace motor 2 with motor 3."

(Teaching data C)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC1, the slope of the relational expression of the rotational speed of the motor shaft with respect to the supplied current is determined so that the response characteristics are as described above. A drone was provisionally designed to use motor 1 with A=A1 (<A2<A3), and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr120 (>Tr100).
The motor 1 has a structure in which the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC1, and the performance is such that the time TR until the roll angle to the right reaches r1 is within Tr100. It was found that it does not have
As a result of replacing motor 1 with motor 3 and conducting a limited flight test again, the time taken until the roll angle to the right reached r1 was TR=Tr099 (<Tr100). Therefore, the data regarding improvement are "improvement required" and "replace motor 1 with motor 3."

From the above, the input data of the teacher data C is the propeller length LLx = LL1, the distance between the propeller rotation center and the body's center of gravity position LCx = LC1, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 2. The slope A=A1.
The output data of the teacher data C are "Improvement required" and "Replace motor 1 with motor 3."

(Teaching data D)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC2 (>LC1), the rotational speed of the motor shaft with respect to the supplied current is adjusted so that the response characteristics are as described above. A drone was provisionally designed to use the motor 3 with the slope of the relational expression A=A3, and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr90 (<Tr100).
The motor 3 has a structure in which the length of the propeller is LLx=LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx=LC2 (>LC1), and the time TR until the roll angle to the right reaches r1 is Tr100. It was found that it has the performance that can be achieved within Therefore, the data regarding improvement is "no improvement required".

From the above, the input data of the teacher data D are the propeller length LLx = LL1, the distance between the propeller rotation center and the body's center of gravity position LCx = LC2, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 3. The slope A=A3.
The output data of the teacher data D is "no improvement required".

(Teaching data E)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC2 (>LC1), the rotational speed of the motor shaft with respect to the supplied current is adjusted so that the response characteristics are as described above. A drone was provisionally designed to use a motor 2 with a slope of the relational expression A=A2 (<A3), and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr95 (<Tr100).
For motor 2, the time TR until the roll angle to the right reaches r1 in a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC2 (>LC1). It was found that it has a performance that can be achieved within Tr100. Therefore, the data regarding improvement is "no improvement required".

From the above, the input data of the teacher data E are the propeller length LLx = LL1, the distance between the propeller rotation center and the body's center of gravity position LCx = LC2, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 3. The slope A=A2.
The output data of the teacher data E is "no improvement required".

(Teaching data F)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body is LCx = LC2 (>LC1), the rotational speed of the motor shaft with respect to the supplied current is adjusted so that the response characteristics are as described above. A drone was provisionally designed to use the motor 1 with the slope of the relational expression A=A1 (<A3), and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr105 (>Tr100).
The motor 1 has a structure in which the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC2 (>LC1), and the time TR until the roll angle to the right reaches r1 is Tr100. It was found that it did not have the performance that could be achieved within the same period. As a result of replacing motor 1 with motor 3 and conducting a limited flight test again, the time taken until the roll angle to the right reached r1 was TR=Tr90 (<Tr100). The data related to improvement are "improvement required" and "replace motor 1 with motor 3."

From the above, the input data of the teacher data F are the length of the propeller LLx = LL1, the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC2, and the relational expression of the rotational speed of the motor shaft with respect to the applied voltage of the motor 3. The slope A=A1.
The output data of the teacher data F are "Improvement required" and "Replace motor 1 with motor 3."

(Teaching data G)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL2 (>LL1) and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC1, the rotational speed of the motor shaft with respect to the supplied current is adjusted so that the response characteristics are as described above. A drone was provisionally designed to use the motor 3 with the slope of the relational expression A=A3, and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr105 (>Tr100).
The motor 3 has a structure in which the length of the propeller is LLx = LL2 (>LL1) and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC1, and the time TR until the roll angle to the right reaches r1 is Tr100. The propeller was too long.
As a result of changing the propeller to a propeller with a length of LL1 and conducting a limited flight test again, the time required for the roll angle to the right to reach r1 was TR=Tr99 (<Tr100). The data regarding improvements are "improvement required" and "propeller changed to a propeller with length LL1".

From the above, the input data of the teacher data G is the propeller length LLx = LL2, the distance between the propeller rotation center and the center of gravity position of the main body LCx = LC1, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 3. The slope A=A3.
The output data of the teacher data G are "Improvement required" and "Propeller changed to a propeller with length LL1".

(Teaching data H)
The predetermined response characteristic is the same as the predetermined response characteristic in the teaching data A described above.
In a structure where the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC2 (<LC1), the rotational speed of the motor shaft with respect to the supplied current is adjusted so that the response characteristics are as described above. A drone was provisionally designed to use the motor 3 with the slope of the relational expression A=A3, and a limited flight test was conducted on the drone manufactured according to the provisional design.

As a result, the time TR until the rightward roll angle reached r1 was Tr105 (>Tr100).
The motor 3 has a structure in which the length of the propeller is LLx = LL1 and the distance between the center of rotation of the propeller and the center of gravity of the main body LCx = LC2 (<LC1), and the time TR until the roll angle to the right reaches r1 is Tr100. The distance LCx between the center of rotation of the propeller and the center of gravity of the main body was too short. As a result of changing the propeller to a position where the distance LCx from the center of gravity of the main body is LC1 (>LC2) and conducting a limited flight test again, the time taken until the roll angle to the right becomes r1 TR = Tr99 (<Tr100 ). The data related to improvement are "improvement required" and "change the propeller to a position where the distance LCx from the center of gravity of the main body is LC1 (>LC2)".

From the above, the input data of the teacher data H is the propeller length LLx = LL2, the distance between the propeller rotation center and the center of gravity position of the main body LCx = LC2, and the relational expression of the rotation speed of the motor shaft with respect to the applied voltage of the motor 3. The slope A=A3.
The output data of the teacher data H are "Improvement required" and "Change the propeller to a position where the distance LCx from the center of gravity of the main body is LC1 (>LC2)".

The teacher data A to H are examples, and the teacher data is not limited to the teacher data A to H, and a large number of teacher data are obtained by conducting various other limited flight tests. The response characteristics are also not limited to the time TR required for the rightward roll angle to reach r1 to be Tr100 (seconds), and training data are obtained with various other response characteristics.

As explained above, in the second embodiment, it is possible to specify data related to improvement of the drone, and the drone can be designed efficiently and the drone can be designed in a short time. In addition, this embodiment has the same effects as the first embodiment.

In the second embodiment, the learned model is used to identify data related to improvement of the drone, but the technology of the present disclosure is not limited to this. For example, a plurality of combinations of the ideal moving state of the drone 10, the actual moving state of the drone 10, and data regarding improvement of the drone are stored in the secondary storage device 222. The specifying unit 516 generates data regarding improvement of the drone based on the plurality of combinations stored in the secondary storage device 222 and the ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the signal. Identify. More specifically, the specifying unit 516 selects a combination that corresponds to the ideal movement state of the drone 10 and the actual movement state of the drone 10 according to the signal from among the plurality of combinations stored in the secondary storage device 222. Drones Identify data for drone improvements. The display unit 514 displays data regarding improvements to the identified drone on the display 224. In this way, data regarding improvements to the drone can be identified. In particular, in this example, model learning can be made unnecessary.

Next, a modification of the aircraft holding device 150 will be described. Since the modified example described below has the same configuration as the embodiment described above, the same configurations will be given the same reference numerals, the explanation thereof will be omitted, and the different parts will be mainly explained.

(First modification)
FIG. 13 shows a schematic configuration of an aircraft holding device 150C1 of a first modification. FIG. 14 shows the drone 10 of the first modification example rising while being held on the mounting table 114.
As shown in FIG. 13, in the aircraft holding device 150C1, the base body 118, support columns 120N1 to 120N4, and casters 122N1 to 122N4 (see also FIG. 2) of the aircraft holding device 150 of the embodiment described above are omitted.

The aircraft holding device 150C1 includes a support table 324 that supports the mounting table 114 and has substantially the same shape and size as the mounting table 114, and supports each corner of the support table 324 at one end and supports the aircraft holding device 150C1 at the other end. It includes four support columns 320N1 to 320N4 that are in contact with the installation location, and a column 330 that supports the mounting table 114.

The pillar 330 includes an outer cylinder 330N1 that supports the support stand 324, and an inner pillar 330N2. A first disc 330N23 (see FIG. 14) is provided at the lower end of the internal column 330N2. A second disc 330N22 is provided on the internal column 330N2 at a predetermined distance above the first disc 330N23. The first disc 330N23 and the second disc 330N22 are discs with a first radius. Note that the radius of the internal cross section of the outer cylindrical body 330N1 is slightly longer (that is, a predetermined length) than the first radius. Therefore, the internal column 330N2 can move up and down inside the outer cylindrical body 330N1.

An opening 326 with a second radius smaller than the first radius is formed in the support base 324 so that the internal column 330N2 can rise and fall.

One end of the internal column 330N2 of the column 330 and the mounting table 114 are connected by a universal joint 112 so that the platform 114 can rotate three-dimensionally based on one end (i.e., the upper end) of the internal column 330N2 of the column 330. has been done. Note that a ball joint may be used instead of the universal joint.

Note that the configuration and operation of each of the drone 10 and the information output device 170 are also the same as in the embodiment described above, so a description thereof will be omitted.

As shown in FIG. 14, when the drone 10 rises while being held on the mounting base 114, the internal column 330N2 of the support 330 connected to the mounting base 114 by the universal joint 112 opens the opening 326 of the support base 324. rise through. The first radius of the second disk 330N22 provided on the lower end side of the internal column 330N2 is larger than the second radius of the opening 326 of the support base 324. Therefore, when the internal column 330N2 rises, the second disk 330N22 hits the periphery of the opening 326 of the support base 324, and the internal column 330N2 cannot rise thereafter. Therefore, the second disk 330N22 has the role of a stopper for the rise of the internal column 330N2.

Furthermore, since a first disk 330N23 and a second disk 330N22 each having a first radius are provided on the lower end side of the internal column 330N2, the internal column 330N2 moves up inside the outer cylinder 330N1. And when descending, the internal column 330N2 can be prevented from inclining.

(Second modification)
FIG. 15 shows a schematic configuration of an aircraft holding device 150C2 of a second modification. FIG. 16 shows how the drone 10 rises slightly (ie, by a predetermined length) from the state where it is held on the mounting table 114, and the support columns 420N1 to 420N4 fall down.
As shown in FIG. 15, the aircraft holding device 150C2 includes a non-extendable support column 440 instead of the extensible support column 110 of the embodiment described above. The aircraft holding device 150C2 includes four support columns 420N1 to 420N4 connected to the base body 118 via universal joints 420J1 to 420J4, instead of the support columns 120N1 to 120N4 fixed to the base body 118 of the embodiment described above. It is equipped with Each corner of the mounting table 114 is supported at the upper end of the support columns 420N1 to 420N4. In the aircraft holding device 150C2, the casters 122N1 to 122N4 (see also FIG. 2) of the embodiment described above are omitted. Note that casters 122N1 to 122N4 may be provided.

One end (i.e., the upper end) of the support column 440 and the mounting table 114 are connected by the universal joint 112 so that the mounting table 114 can rotate three-dimensionally based on one end (i.e., the upper end) of the support column 440. There is. The lower end of the support column 440 is fixed to the base body 118.

As described above, since the non-extendable support column 440 and the mounting table 114 are connected by the universal joint 112, the drone 10 can rotate three-dimensionally while being held on the mounting table 114. cannot rise.

Note that the configuration and operation of the drone 10 are also substantially the same as in the embodiment described above, so a detailed explanation thereof will be omitted, and the different parts will be explained based on acquired information.

First, as shown in FIG. 15, the operator holds the drone 10 on the mounting table 114 and supports the mounting table 114 by the support columns 420N1 to 420N4. Next, the operator operates the operating device 50 to raise the drone 10. One end of the support column 440 and the mounting table 114 are connected by the universal joint 112, but due to the structure of the universal joint 112, the drone 10 rises slightly (that is, by a predetermined length).

When the drone 10 rises slightly (ie, by a predetermined length), the support columns 420N1 to 420N4 fall down, as shown in FIG. 16. Therefore, the drone 10 does not rise while being held on the mounting table 114, but can rotate three-dimensionally.

The configuration of the information output device 170 is the same as that of the embodiment described above, so a description thereof will be omitted. The operation of the information output device 170 is the same as that of the embodiment described above, except that the position of the drone 10 is not calculated or detected, so a description thereof will be omitted.

In the first and second embodiments described above, the response characteristics of each of a plurality of types of drones having different response characteristics are stored in the ROM 214 or the secondary storage device 222 as data indicating the type of drone. Correspondingly, the ideal position and attitude of the drone 10 are stored and estimated, and the actual position and attitude of the drone 10 are calculated, but the present embodiment is not limited thereto. For example, the direction of movement of the drone (ideal direction) is specified from the instruction signal from the operating device 50, the direction of movement of the drone (actual direction) is calculated from the image obtained by the camera, and the ideal direction is determined. and the actual direction may be stored or displayed as information for specifying data related to a flight failure of the drone 10. A value indicating the degree of coincidence between the ideal direction and the actual direction may be calculated, and the value indicating the degree of coincidence between the ideal direction and the actual direction may be stored or displayed. Therefore, even if the ROM 214 or the secondary storage device 222 does not store the response characteristics of each of a plurality of types of drones having different response characteristics, it is possible to provide information for identifying data related to flight failures of the drone 10. Can be done.

In each of the examples described above, a cable tie with which the operator ties the support part 20 of the drone 10 to the mounting table 114 is used as the holding part 116. However, the technology of the present disclosure is not limited to this, and the support part 116 of the drone 10 is 20 and the mounting table 114 may be connected and disconnected automatically. For example, through holes are formed at the tips of the four supporting parts 20. Further, by rotation of the motor, a rod (for example, a pinion) is inserted into the through hole at the tip of each support part 20 of the drone 10 placed on the mounting table 114 via a moving mechanism such as a rack and pinion mechanism. and detach. Thereby, the support part 20 of the drone 10 and the mounting table 114 are connected and released. In addition to the rack and pinion mechanism, the moving mechanism may include a rod that is biased toward the connection side or the detachment side, and an eccentric cam that moves the rod between the detachment side and the detachment side. .

As described above, the operator views at least one of the screen 224S of the display 224 (see FIG. 8) and the graphs (FIGS. 9 and 10) that display information for evaluating the response characteristics of the drone 10. The actual response characteristics of the drone 10 can be confirmed. If the operator determines that the actual response characteristics of the drone 10 are not worse than the response characteristics of the above-mentioned stored climbing speed of the type of the drone 10 and that it will not fall even if a free flight test is performed, the operator determines that the support part of the drone 10 20 and the mounting table 114 is instructed to the computer 210 (see FIG. 6A) of the calculation storage device 200 of the information output device 170 via the input device 228. Thereby, the connection between the support part 20 and the mounting table 114 of the drone 10 is released, and a free flight test can be performed on the drone 10.

In the embodiment and each modification example described above, a restricted flight test is performed in order to obtain teacher data, but a free flight test may also be performed.

Although the embodiment and each modification example described above has been described using the drone 10, the technology of the present disclosure is not limited thereto. For example, instead of a drone, other unmanned aircraft, such as a radio-controlled airplane and a radio-controlled unmanned helicopter, or even a manned aircraft, such as a radio-controlled helicopter that can carry a person, may be used. good.

In the present disclosure, only one or two or more of each component (device, etc.) may exist as long as there is no contradiction.

In each of the examples described above, the information output processing is realized by a software configuration using a computer, but the technology of the present disclosure is not limited to this. For example, instead of a software configuration using a computer, information output processing may be performed only by a hardware configuration such as an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). . Part of the information output processing may be executed by a software configuration, and the remaining processes may be executed by a hardware configuration.

Note that the programs described above can be stored and provided to a computer using various types of non-transitory computer-readable media. Non-transitory computer-readable media includes various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, and CDs. - R/W, semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). The program may also be provided to the computer on various types of temporary computer-readable media. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can provide the program to the computer via wired communication channels, such as electrical wires and fiber optics, or wireless communication channels.

The information output processing described above is just an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be changed within the scope of the main idea.

All documents, patent applications, and technical standards mentioned herein are incorporated by reference, as if each individual document, patent application, and technical standard was specifically and individually indicated to be incorporated by reference. Incorporated herein by reference.

(Additional note)
Based on the above disclosure content, the following additional notes are proposed.
(Additional note 1)
an aircraft holding device that holds the mounted aircraft;
a specific device that specifies data related to the performance of an aircraft held movably in three dimensions in the aircraft holding device;
A specific system comprising:
The aircraft holding device includes:
A mounting table and
a holding part that holds the aircraft placed on the mounting stand;
A pillar that supports the aforementioned mounting stand;
a connecting portion that connects one end of the pillar and the mounting base so that the mounting base is three-dimensionally movable with respect to the one end of the pillar;
Equipped with
The aircraft receives a signal indicating the content of movement from an instruction device instructing movement, and moves in accordance with the received signal indicating the content of movement,
The specific device is
a receiving unit that receives the signal;
regarding the performance of the aircraft based on the ideal movement state of the aircraft according to the received signal and the actual movement state of the aircraft moved while being held movably in three dimensions by the aircraft holding device. A specific part that specifies data;
Equipped with
Specific system.

(Additional note 2)
The aircraft holding device further includes a base body having a plurality of moving members on a lower surface,
the other end of the support is attached to the top surface of the base;
Specific system in Appendix 1.

(Additional note 3)
identifying data regarding the performance of the aircraft using a trained model used to receive a signal indicating the content of the movement from the indicating device and identifying data regarding the performance of the aircraft moving in response to the received signal; A recording medium that records a program that causes a computer to execute a process,
The performance state determination process includes:
obtaining an ideal movement state of the aircraft and an actual movement state of the aircraft in response to the signal;
identifying data regarding the performance of the aircraft based on the trained model, the obtained ideal movement state of the aircraft, and the actual movement state of the aircraft;
recording media, including

(Additional note 4)
A recording medium that stores a program that causes a computer to execute a learning process that generates a learned model by learning a model that specifies data related to aircraft performance,
The learning process is
obtaining training data having input information for specifying data regarding the performance of the aircraft and outputting data regarding the performance of the aircraft;
using the training data to learn a model whose input is information for specifying data related to the performance of the aircraft and whose output is data related to the performance of the aircraft;
recording media, including

(Appendix 5)
A recording medium that records a trained model for identifying data regarding aircraft performance,
The trained model is trained using training data that inputs information for specifying data regarding the performance of the aircraft and outputs data regarding the performance of the aircraft,
The learned model receives as input information for identifying data regarding the performance of the aircraft, and identifies data regarding the performance of the aircraft based on the received information for identifying data regarding the performance of the aircraft. have a computer perform a process,
recoding media.

Claims

A specific device that receives a signal indicating the content of movement from an instruction device and specifies data regarding the performance of a moving aircraft according to the received signal,
a receiving unit that receives the signal;
an identification unit that identifies data regarding the performance of the aircraft based on an ideal movement state of the aircraft and an actual movement state of the aircraft according to the received signal;
A specific device equipped with
further comprising a storage unit that stores a plurality of combinations of an ideal movement state of the aircraft, an actual movement state of the aircraft, and data regarding the performance of the aircraft,
The determination unit determines the performance of the aircraft based on the plurality of combinations stored in the storage unit and the ideal movement state of the aircraft and the actual movement state of the aircraft according to the received signal. identify data,
The identification device according to claim 1.
further comprising a storage unit that stores a trained model trained using teacher data in which an ideal movement state of the aircraft and an actual movement state of the aircraft are input, and data regarding the performance of the aircraft is output;
The determination unit determines the performance of the aircraft based on the learned model stored in the storage unit, an ideal movement state of the aircraft and an actual movement state of the aircraft according to the received signal. identify data,
The identification device according to claim 1.
At least one of the aircraft and the specific device includes a detection unit that detects the actual movement state of the aircraft,
The actual movement state of the aircraft used by the determination unit is the actual movement state detected by the detection unit.
The identification device according to any one of claims 1 to 3.
The identification device according to any one of claims 1 to 4, wherein the aircraft moves while being held in a three-dimensionally movable manner by an aircraft holding device.
Data regarding the performance of the aircraft may include data regarding improvements to the aircraft at the stage of designing it to have predetermined response characteristics or failures of the aircraft manufactured according to a completed design to have predetermined response characteristics. Data regarding
The identification device according to any one of claims 1 to 5.
identifying data regarding the performance of the aircraft using a trained model used to receive a signal indicating the content of the movement from the indicating device and identifying data regarding the performance of the aircraft moving in response to the received signal; A program that causes a computer to execute a process,
The specific processing is
obtaining an ideal movement state of the aircraft and an actual movement state of the aircraft in response to the signal;
identifying data regarding the performance of the aircraft based on the trained model, the obtained ideal movement state of the aircraft, and the actual movement state of the aircraft;
programs, including.
A learning method that generates a trained model by learning a model that specifies data regarding aircraft performance, the method comprising:
obtaining training data having input information for specifying data regarding the performance of the aircraft and outputting data regarding the performance of the aircraft;
using the training data to learn a model whose input is information for specifying data related to the performance of the aircraft and whose output is data related to the performance of the aircraft;
Learning methods including.
A learning device that generates a trained model by learning a model that specifies data regarding aircraft performance,
an acquisition unit that acquires training data having input information for specifying data regarding the performance of the aircraft and outputting data regarding the performance of the aircraft;
a learning processing unit that uses the training data to learn a model that receives information for specifying data regarding the performance of the aircraft as input and outputs data regarding the performance of the aircraft;
learning devices including;
A program that causes a computer to execute a learning process to generate a learned model by learning a model that specifies data related to aircraft performance, the program comprising:
The learning process is
obtaining training data having input information for specifying data regarding the performance of the aircraft and outputting data regarding the performance of the aircraft;
using the training data to learn a model whose input is information for specifying data related to the performance of the aircraft and whose output is data related to the performance of the aircraft;
programs containing.
A trained model for identifying data regarding aircraft performance, the model comprising:
The trained model is trained using training data that inputs information for specifying data regarding the performance of the aircraft and outputs data regarding the performance of the aircraft,
The learned model receives as input information for identifying data regarding the performance of the aircraft, and identifies data regarding the performance of the aircraft based on the received information for identifying the data regarding the performance of the aircraft. have a computer perform a process,
Trained model.