JP2020057944A - Status notification device of part of unmanned flight device, status notification method of part of unmanned flight device, and program - Google Patents

Abstract

To provide a status notification device of a part of an unmanned flight device, a status notification method of a part of an unmanned flight device, and a program.SOLUTION: A status notification device 300 of a part of an unmanned flight device 1 includes learning result data storage means that stores learning result data obtained by machine-learning a relationship between a plurality of different use conditions of the unmanned flight device 1 and durability of a part of the unmanned flight device 1, schedule condition receiving means that receives schedule condition data indicating planned usage conditions of a specific unmanned flight device, state determination means that determines a state of a part of the specific unmanned flight device on the basis of the learning result data, the scheduled condition data, and passage of time, and notification sending means that sends a notification to a user of the specific unmanned flight device as to the status of the part of the specific unmanned flight device.SELECTED DRAWING: Figure 12

Description

  The present invention relates to an unmanned aerial vehicle component status notification device, an unmanned aerial vehicle component status notification method, and a program.

  Conventionally, the use of small unmanned aerial vehicles (also called "drones") has been proposed. A technique for acquiring video information using such a drone has been proposed (for example, Patent Document 1).

JP 2006-27331 A

  By the way, parts such as a drone motor and a battery for supplying power to the motor are consumables. For example, if a drone is flown when the life of the motor is near, the motor may break down during flight and the drone may crash.

  An object of the present invention is to solve such a problem, and an object of the present invention is to provide a component status notification device for an unmanned aerial vehicle, a component status notification method and a program for an unmanned aerial vehicle.

  A first invention is a learning result data storage unit that stores learning result data obtained by machine learning a relationship between a plurality of different use conditions of an unmanned aerial vehicle and durability of parts of the unmanned aerial vehicle, Scheduled condition receiving means for receiving scheduled condition data indicating a scheduled use condition of the unmanned aerial vehicle; and a state of parts of the specific unmanned aerial vehicle based on the learning result data, the scheduled condition data, and time lapse. State notification means for transmitting a notification regarding the state of the component of the specific unmanned aerial vehicle to a user of the specific unmanned aerial vehicle, and a state notification of the component of the unmanned aerial vehicle. Device.

  According to the configuration of the first invention, the state notification device for the component of the unmanned aerial vehicle stores the learning result data obtained by machine learning the relationship between the use condition and the durability of the component, The means can determine the state of the part based on the scheduled condition data and the passage of time, and the notification transmitting means can transmit a notification regarding the state of the part to the user of the specific unmanned aerial vehicle. .

  The second invention is a second learning result data storage unit that stores second learning result data obtained by machine learning the output of the parts and the state of the parts of the unmanned flying device, and a part of the specific unmanned flying device Output data receiving means for receiving output data indicating the output of the second learning result data and, based on the output data, a second state determination means for determining the state of the component of the specific unmanned flying device, An unmanned aerial vehicle component status notification device, comprising: a second notification transmission unit configured to transmit a notification regarding a status of the specific unmanned aerial vehicle component to a specific unmanned aerial vehicle user.

  According to the configuration of the second invention, since the learning result data is data obtained by machine learning of the output of the unmanned aerial vehicle, it is highly reliable data reflecting the actual output. Then, the second state determination means can determine the state of the highly reliable component based on the second learning result data and the output data of the specific unmanned aerial vehicle, and the second notification transmission means It is possible to transmit a notification regarding the state of a component having a high degree.

  A third aspect of the present invention is the component status notification device of the unmanned aerial vehicle, comprising a second learning unit configured to machine-learn the output and generate the second learning result data in the configuration of the second invention.

  According to the configuration of the third invention, the component state notification device itself of the unmanned aerial vehicle can generate the second learning result data.

  In a fourth aspect based on the configuration of the third aspect, based on the specific output, state data indicating a state of the component is generated, and learning data including the output data and the state data is generated. The second learning means is a component status notification device of an unmanned aerial vehicle, configured to learn the learning data.

  According to the configuration of the fourth aspect of the invention, the component status notification device of the unmanned aerial vehicle can generate learning data that is a prerequisite for machine learning. The state data is generated based on a specific output, but the output included in the learning data is not limited to the specific output. For this reason, the learning data is composed of various outputs and state data.

  In a fifth aspect based on the configuration of the third aspect or the fourth aspect, the second state determining means is configured to determine a future state of the component, and the second notification The means is configured to notify a user of the particular unmanned aerial vehicle of a current state of the part and to notify a future state of the part, wherein a state of the part of the unmanned aerial vehicle is configured. Notification device.

  According to the configuration of the fifth aspect, the future state of the component can be notified.

  In a sixth aspect based on the configuration according to any one of the third to fifth aspects, the second learning means machine-learns the learning data that changes with the passage of time, and determines a future state of the component. An unmanned aerial vehicle component status notification device configured to generate the second learning result data that can be determined.

  The seventh invention is the configuration according to any one of the second invention to the sixth invention, wherein the set of outputs received when generating the learning data is applied to the second result data, If there is a difference between the state data included in the data and the state of the component determined by applying the second result data, the second result data is adjusted to enhance the difference. It is a status notification device for parts of an unmanned aerial vehicle, which has a learning means.

  An eighth invention provides an unmanned flight having learning result data storage means for storing learning result data obtained by machine learning a relationship between a plurality of different use conditions of an unmanned aerial vehicle and the durability of parts of the unmanned aerial vehicle. The state notification device of the parts of the device, the scheduled condition receiving step of receiving scheduled condition data indicating a scheduled use condition of the specific unmanned flying device, and the learning result data, the scheduled condition data, and based on the elapsed time. A state determining step of determining a state of a part of the specific unmanned aerial vehicle, a notification transmitting step of transmitting a notification regarding a state of the part of the specific unmanned aerial vehicle to a user of the specific unmanned aerial vehicle; , A method of notifying the status of parts of an unmanned aerial vehicle.

  A ninth invention is a component status notification device of an unmanned aerial vehicle that has a second learning result data storage unit that stores second learning result data acquired by machine learning of output information of a component of an unmanned aerial vehicle, An output data receiving step of receiving output data indicating an output of a specific unmanned aerial vehicle, and a second state for predicting a state of a component of the specific unmanned aerial vehicle based on the second learning result data and the output data A method of notifying a status of a part of an unmanned aerial vehicle that performs a determination step and a second notification transmitting step of transmitting a notification regarding a state of a part of the specific unmanned aerial vehicle to a user of the specific unmanned aerial vehicle It is.

  A tenth invention is a computer that controls a state notification device for parts of an unmanned aerial vehicle, obtained by machine learning the relationship between a plurality of different use conditions of the unmanned aerial vehicle and the durability of the parts of the unmanned aerial vehicle. Learning result data storage means for storing learning result data; scheduled condition receiving means for receiving scheduled condition data indicating a scheduled use condition of a specific unmanned aerial vehicle; the learning result data; the scheduled condition data; State determining means for determining a state of a part of the specific unmanned aerial vehicle based on the notification transmitting means for transmitting a notification regarding a state of the part of the specific unmanned aerial vehicle to a user of the specific unmanned aerial vehicle , Is a program to function as.

  An eleventh invention is directed to a computer for controlling a state notification device for parts of an unmanned aerial vehicle, a second learning result for storing second learning result data obtained by machine learning of output information of parts for an unmanned aerial vehicle. A data storage unit, an environmental condition data receiving unit that receives output data indicating an output of the specific unmanned aerial vehicle, and predicts a state of a part of the specific unmanned aerial vehicle based on the second learning result data and the output data. A program for causing the user of the specific unmanned aerial vehicle to function as a second notification transmitting unit that transmits a notification regarding a state of a part of the specific unmanned aerial vehicle to the user of the specific unmanned aerial vehicle.

  As described above, according to the present invention, it is possible to notify the status of components of an unmanned aerial vehicle.

It is a schematic diagram showing an operation of a first embodiment of the present invention. It is the schematic which shows an example of the data for learning. It is a schematic diagram showing an example of learning result data. It is the schematic which shows the structure of an unmanned flight device. It is the schematic which shows the function block of an unmanned flight device. FIG. 3 is a schematic diagram illustrating functional blocks of a base station. It is a schematic flowchart which shows an example of the process of machine learning. It is the schematic which shows an example of schedule condition data. 5 is a schematic flowchart illustrating an operation of a base station. It is a schematic diagram showing an operation of a second embodiment of the present invention. It is the schematic which shows an example of the data for learning. It is the schematic which shows an example of 2nd learning result data. 5 is a schematic flowchart illustrating an operation of a base station.

<First embodiment>
Hereinafter, embodiments for implementing the present invention (hereinafter, embodiments) will be described in detail. In the following description, the same components are denoted by the same reference numerals, and description thereof will be omitted or simplified. The description of the configuration that can be appropriately performed by those skilled in the art will be omitted, and only the basic configuration of the present invention will be described.

  The unmanned aerial vehicles 1A to 1H shown in FIG. 1 are unmanned aerial vehicles that obtain thrust by rotating a propeller and fly autonomously on a predetermined path, and are examples of an unmanned aerial vehicle. The unmanned aerial vehicles 1A to 1H are collectively referred to as “unmanned aerial vehicles 1”.

  The base station 300 has a computer as a component. The computer determines the state of the parts of the unmanned aerial vehicle 1 and sends a notification regarding the state of the parts to a user (not shown) of the unmanned aerial vehicle 1. The state of a part is, for example, a state in which a specific part or a part of a specific type has a short product life, has a defect, or is likely to have a defect. The “specific component” is a component in a case where there are a plurality of components having the same specification and the individual components are distinguished. The “part of a specific type” is a part in a case where there are a plurality of parts having the same specification and the individual parts are not distinguished. Note that in this specification, “send to user” or “receive from user” means communicating with a communication device such as a computer or a portable terminal managed by the user.

  As will be described later, the drone 1 is composed of various components. These parts are consumables, and the product life to function within a given specification depends on the durability of the parts. The durability varies depending on the application in which the unmanned aerial vehicle 1 is used, the environment such as temperature and humidity, and the conditions such as frequency of use. In the present specification, this condition is referred to as “usage condition”.

  The base station 300 stores learning result data obtained by machine learning the relationship between a plurality of different use conditions of the unmanned aerial vehicle (not limited to the unmanned aerial vehicle 1 but many and many unmanned aerial vehicles) and the durability of the parts. ing. The learning result data need not be generated by the base station 300, and may be generated by a computer other than the base station 300. However, in the present embodiment, the description will be given on the assumption that the base station 300 generates the learning result data.

  FIG. 2 is a schematic diagram illustrating learning data used by the base station 300 for machine learning. The usage conditions include usages such as photographing and pesticide spraying, usage locations such as mountains and the sea (natural conditions Nx), usage regions such as northern Japan, southern Japan, and Kanto (natural conditions Mx), and usage frequencies. You. Durability is indicated as a replacement period, for example, as an average operating time of the drone before replacement of the battery, motor, and propeller is required for each use condition. For example, if the usage is shooting, the usage area is a mountainous area, the usage frequency is 10 hours / week, and the usage area is northern Japan, the durability of the battery is 100 hours on average, the durability of the motor is 10,000 hours on average, The durability of the propeller averages 20,000 hours.

  The base station 300 machine-learns the learning data to generate learning result data shown in FIG. Based on the learning result data, the scheduled condition data indicating the scheduled use conditions of the unmanned aerial vehicle 1, and the elapsed time, for example, the base station 300 provides a specific component (for example, , A battery) (for example, a state that the product life is near).

  Hereinafter, the configuration of the unmanned aerial vehicle 1 will be described with reference to FIG. As shown in FIG. 4, the drone 1 has a housing 2. The housing 2 includes an autonomous flight device including a computer, a wireless communication device, a positioning device using positioning radio waves from a navigation satellite system such as a GPS (Global Positioning System), a temperature sensor, a humidity sensor, and a battery. Is arranged. The autonomous flight device includes an inertial sensor, an electronic compass, and a barometric pressure sensor. Further, a camera 14 is disposed on the housing 2 via a fixing device 12.

  The drone 1 acquires an external image by the camera 14. The camera 14 is a visible light camera, but is not limited thereto, and may be, for example, a multispectral camera. The fixing device 12 is a three-axis fixing device (so-called gimbal) capable of minimizing blurring of an image captured by the camera 14 and controlling the optical axis of the camera 14 in an arbitrary direction.

  An antenna 16 for receiving positioning radio waves is disposed above the housing 2. The housing 2 also has an antenna (not shown) for receiving a control signal from a control device (called a “propo”), and an antenna (not shown) for transmitting an image captured by the camera 14 to the outside. (Not shown).

  A round bar-shaped arm 4 is connected to the housing 2. A motor 6 is connected to each arm 4, and a propeller 8 is connected to each motor 6. Each motor 6 is a DC motor (brushless DC motor). The rotation speed of each motor 6 is controlled by a motor driver (not shown). The motor drivers are independently controlled by the autonomous flight devices. The motors 6 are controlled independently of each other, so that the drone 1 can be freely moved vertically and horizontally, stopped (hovered) in the air, and controlled in attitude.

  A protective frame 10 is connected to the arm 4 to prevent the propeller 8 from directly contacting an external object. The arm 4 and the protection frame 10 are formed of, for example, carbon fiber reinforced plastic, and are configured to be lightweight while maintaining strength. In addition, the protection frame 10 is not arrange | positioned at the drones 1E-1H among the drones 1 of FIG.

  FIG. 5 is a diagram illustrating a functional configuration of the drone 1. The unmanned aerial vehicle 1 includes a CPU (Central Processing Unit) 50, a storage unit 52, a wireless communication unit 54, a satellite positioning unit 56, an inertial sensor unit 58, a drive control unit 60, an image processing unit 62, and a power supply unit 64.

  The unmanned aerial vehicle 1 can communicate with the outside by the wireless communication unit 54. The unmanned aerial vehicle 1 receives an instruction such as start from the control device (propo) operated by the user via the wireless communication unit 54.

  The drone 1 can measure the position of the drone 1 itself by the satellite positioning unit 56 and the inertial sensor unit 58. The satellite positioning unit 56 basically measures the position of the unmanned aerial vehicle 1 by receiving positioning radio waves from four or more navigation satellites. The satellite positioning unit 56 is an example of positioning means for positioning the current position. The inertial sensor unit 58 measures the position of the unmanned aerial vehicle 1 by accumulating the movement of the unmanned aerial vehicle 1 from the nearest positioning position by the satellite positioning unit 56 using, for example, an acceleration sensor and a gyro sensor. The position information of the drone 1 itself is used for determining a moving route of the drone 1 and for autonomous movement, and is also used for linking image data captured by the camera 14 with coordinates (position).

  By the drive control unit 60, the drone 1 controls the rotation of the propeller 8 (see FIG. 4) connected to each motor 6 (see FIG. 4), and controls the posture such as vertical and horizontal movement, suspension in the air, and inclination. It has become.

  The drone 1 operates the camera 14 (see FIG. 4) to acquire an external image, and processes the acquired image data by the image processing unit 62.

  The power supply unit 64 is, for example, a replaceable rechargeable battery, and supplies power to each unit of the unmanned aerial vehicle 1.

  The storage unit 52 stores various data and programs necessary for autonomous movement, such as data indicating a movement plan for autonomous movement from a starting point to a destination position.

  FIG. 6 is a diagram illustrating a functional configuration of a computer that is a component of the base station 300. The base station 300 has a CPU 302, a storage unit 304, a wireless communication unit 306, and a power supply unit 308.

  The base station 300 can communicate with the outside by the wireless communication unit 306. The base station 300 communicates with the user of the drone 1 by the wireless communication unit 306.

  The storage unit 304 stores various data and programs for operating the computer. The storage unit 304 also stores a learning program, a scheduled condition receiving program, a state determination program, and a notification program. The CPU 302 and the learning program are examples of learning means. The CPU 302 and the scheduled condition receiving program are an example of scheduled condition receiving means. The CPU 302 and the state determination program are examples of a state determination unit. The CPU 302 and the notification program are examples of a notification unit.

  The base station 300 learns, for example, the learning data shown in FIG. 2 by a learning program, and generates learning result data shown in FIG. The base station 300 machine-learns using various usage conditions of the unmanned aerial vehicle (not limited to the unmanned aerial vehicle 1) and a large number of data indicating the durability of the component as learning data, and performs the component learning based on the usage condition of the unmanned aerial vehicle. Learning result data, which is a neural network model for determining durability, is generated and stored in the storage unit 304.

  An outline of a method of generating learning result data (machine learning process) will be described with reference to FIG. When the learning data is input to the model of the neural network of the computer (step S1 in FIG. 7), learning is performed by a machine learning algorithm (step S2). The neural network model is, for example, a neural network conceptually shown in FIG. The neural network includes an input layer, one or more hidden layers, and an output layer. FIG. 3 shows only one hidden layer, but there are actually a plurality of hidden layers. That is, the learning according to the present embodiment is machine learning (deep learning) using a multilayer neural network (Deep Neural Network). For example, weights such as weights w1a and w2a are set for input values input to the hidden layer and the output layer. In the hidden layer and the output layer, for example, a bias such as B1 (threshold in each layer) is set. By adjusting the weights w1a and w2a of the input values input to the hidden layer and the output layer, and the biases (thresholds in each layer) B1 and the like in the hidden layer and the output layer by machine learning, the output from the output layer becomes correct. Try to get closer. The learning is performed on the durability of each part, such as the battery, motor, and propeller.

  As shown in FIG. 2, the learning data is data in which use conditions such as a use, a use place, a use frequency, and a use area are associated with actual durability. For example, regarding the place of use, natural conditions such as temperature and humidity are different depending on the difference between the mountainous area, the sea, and the ground. Further, regarding the use area, natural conditions such as temperature and humidity are different depending on a geographical range such as northern Japan and southern Japan. In step S2 of FIG. 7, a large number of learning data are learned. As a result of the learning, the internal structure of the neural network model changes. Specifically, the values of the weight and the bias are adjusted and change. As a result of the learning, for example, learning result data shown in FIG. 3 is generated (step S3 in FIG. 7). FIG. 3 shows a model for determining the durability of the battery. A similar model is generated for each component. These models constitute Artificial Intelligence (AI) and are used to determine the durability of specific drone 1 components, as described below.

  The base station 300 receives scheduled condition data indicating a scheduled use condition of the specific drone 1 from a user of the specific drone 1 by the scheduled condition receiving program. The contents of the expected use conditions include, for example, the use, the use place, the use area, and the use frequency as shown in FIGS. 8A to 8C.

  The base station 300 determines the state of the component of the specific unmanned aerial vehicle 1 using the state determination program. The determination of the state of the component is performed using the above-described learning result data, that is, the neural network. For example, if the scheduled condition data is the data shown in FIG. 8A, the base station 300 determines that the durability of the battery is 120 hours based on the learning result data of FIG. Then, base station 300 determines the remaining time of durability based on the elapsed time.

  The base station 300 transmits, to the user of the specific drone 1, a notification regarding the state of the components of the specific drone 1 by a notification program. For example, when it is determined that it is within a predetermined time until the replacement time indicated as the durability of the battery of the specific unmanned aerial vehicle 1, it is notified that the battery replacement time is near. The predetermined time is, for example, 24 hours (hour).

  Hereinafter, the processing by the base station 300 will be described with reference to FIG. When the base station 300 receives the input of the scheduled condition data from the user of the specific drone 1 (step ST1 in FIG. 9), the base station 300 applies the scheduled condition data to the learning result data (step ST2). The durability of each component is determined (step ST3). Subsequently, the base station 300 starts time counting (step ST4), and when it is determined that it is time to replace any part within a predetermined time (step ST5), transmits a notification to the user (step ST6). ). The predetermined time is, for example, 24 hours. The base station 300 repeats steps ST4 to ST7 until it is determined that the termination condition is satisfied, for example, when a notification from a specific user that the notification of the component status is unnecessary is not made (step ST7).

<Second embodiment>
The second embodiment will be described focusing on parts different from the first embodiment.
The base station 300A of the second embodiment receives output data indicating the output of a component at predetermined time intervals from a plurality of unmanned aerial vehicles 1A and the like (see FIG. 10). The base station 300A processes the output data by artificial intelligence (AI) constituted by a multilayer neural network (Deep Neural Network), and causes malfunction of parts such as a motor, a battery, and a propeller. It is determined whether there is a defect or a shift in the position of the center of gravity (hereinafter, referred to as “defect”) (hereinafter, referred to as “state determination”). Prior to the state determination, the base station 300A learns the relationship between the output data received from the plurality of drones 1A and the like and the actual malfunction by machine learning using a multilayer neural network. Further, base station 300A generates learning data that is a premise of machine learning using a specific output included in the output data.

  Various data and programs for operating a computer are stored in the storage unit 304 (see FIG. 6) of the base station 300A. The storage unit 304 also stores an output data receiving program, a learning data generation program, a second learning program, a second state determination program, a second notification program, and a reinforcement learning program. The CPU 302 and the output data receiving program are examples of an output data receiving unit. The CPU 302 and the learning data generation program are examples of learning data generation means. The CPU 302 and the second learning program are examples of a second learning unit. The CPU 302 and the second state determination program are an example of a second state determination unit. The CPU 302 and the second notification program are examples of a second notification unit. The CPU 302 and the reinforcement learning program are examples of reinforcement learning means.

  The base station 300A receives output data at predetermined time intervals from the drone 1A or the like by using the output data receiving program. The time interval at which the output data is received may be every few seconds, or may be at a time interval deviated to some extent (for example, every 10 minutes).

  As shown in FIG. 10, the base station 300A communicates with a plurality of unmanned aerial vehicles 1A and the like, and receives output data indicating outputs from components such as the unmanned aerial vehicles 1A illustrated in FIG. 11 for each of the unmanned aerial vehicles 1A and the like. ing.

  The output data is data indicating an output from a component such as the unmanned aerial vehicle 1A. These data are associated with the machine ID for identifying the drone 1A and the like and the time at which the data was received. These outputs include, for example, the number of rotations of the motor, the power consumption, the rotation time (of the motor), the altitude (flight), the temperature output from the temperature sensor, and the remaining battery level, as shown in FIG. Although not shown in FIG. 11, these outputs include, for example, radio wave intensity and positioning position received by one or more positioning devices using positioning radio waves transmitted from a GPS satellite or the like, and a signal from a pressure sensor. Includes output pressure, humidity output from a humidity sensor, output from an electronic compass, output from a three-axis gyro that constitutes an inertial sensor (IMU), motor sound, output from a motor driver, and the like. The output from the motor driver is data indicating the magnitude and direction of the electric power (current and voltage) supplied to the windings (coils) constituting the stator of the motor 6, the number of rotations of the motor 6, and the like.

  The base station 300A generates the learning data shown in FIG. 11 using the learning data generation program. The base station 300A determines the state of each component based on the output data, generates state data, and generates learning data including the output data and the state data.

  The base station 300A determines the state of each component based on a predetermined reference using a specific output. Hereinafter, an example of a method (algorithm) of determining the state of the motor 6 will be described as a method in which the base station 300A determines the state of the component. The base station 300A monitors the power consumption of each motor 6 such as the drone 1A. It is considered that the power consumption of each motor 6 is directly reflected on the malfunction of each motor 6. Specifically, the base station 300A receives and monitors data indicating the current value (I) and the voltage value (V) supplied to the stator of the motor 6, that is, power (W = V × I).

  The base station 300A determines whether or not the actual power consumption of each motor 6 is within an allowable range. For example, the base station 300A compares the power consumption of each motor 6 with respect to the drone 1A that is the specific drone 1 using information indicating the power consumption of each motor 6 in the state of hovering in the air, for example. I do. In the state of the suspension in the air, the performance of each motor 6 is the same, and when the center of gravity of the unmanned aerial vehicle 1A is located at the center between the four motors 6, the power consumption of each motor 6 should be the same. If there is a difference in power consumption between the motors 6, it means that there is an abnormality in any of the motors 6, or that the center of gravity of the unmanned aerial vehicle 1A is shifted from the center position between the four motors 6. The base station 300A determines whether or not the power consumption of one of the four motors 6 is within a predetermined allowable range in comparison with the average value Wab of the power consumption of the other three motors 6, for example. Judge. In the predetermined range, for example, the difference is within 5%. The base station 300A performs this process for the power consumption of each motor 6. For example, if the four motors 6 are motors 6A, 6B, 6C, and 6D, first, the power consumption WA of the motor 6A is compared with the average value Wab of the power consumption of the motors 6B, 6C, and 6D, and (WA-Wab ) / Wab is determined to be within ± (plus / minus) 5%. The same processing is performed with the relation between the power consumption WB of the motor 6B and the average power consumption Wab of the motors 6A, 6C and 6D, the power consumption WC of the motor 6C and the average power consumption Wab of the motors 6A, 6B and 6D. And the relationship between the power consumption WD of the motor 6D and the average value Wab of the power consumption of the motors 6A, 6B and 6C.

  For example, when determining that the power consumption of one motor 6 is not within the allowable range, the base station 300A determines whether the operation of the motor 6 is normal. For example, information on the power supplied to the motor 6 and the rotation speed of the motor 6 at that time is acquired, and the power supplied to the motor 6 and the rotation speed (hereinafter, referred to as “specifications”) stored in the storage unit 304 in advance are stored. This is compared with data indicating the relationship of “scheduled rotation speed”. The base station 300A determines whether the difference between the actual rotation speed and the planned rotation speed is within a predetermined allowable range, for example, within 5%. Base station 300A determines that there is no abnormality in motor 6 if it is within a predetermined allowable range, and determines that motor 6 is abnormal if it is not within the predetermined allowable range. If there is no abnormality in the motor 6, it is determined that a failure other than the failure of the motor 6 has occurred, such as the center of gravity of the unmanned aerial vehicle 1A being shifted from the center position between the four motors 6.

  For other components such as a battery and a propeller, the state is determined based on a predetermined method (the description is omitted), and learning data is generated. As described above, the base station 300A itself generates learning data for machine learning. The base station 300A receives output data at predetermined time intervals from the drone 1A and the like, and updates the learning data.

  As described above, the learning data includes a specific output and state data. Although the output data includes various outputs, for the motor 6, as in the case of generating the status data based on the power consumption, the status data is determined based on the output value directly related to the specific component. Is generated on the basis of the determination method (algorithm). On the other hand, the state of each component may be reflected in a method other than a predetermined determination method (algorithm) or in an output value not used for generating learning data. In addition, there is a case where a transition from a normal state to a failure occurs with the passage of time. In this regard, as described below, the base station 300A learns various outputs included in the learning data and transitions of the relationship between the output and the state data with the passage of time, and specifies a specific component or a specific type. A neural network (second learning result data) for determining the current and future state of the part is generated.

  The base station 300A machine-learns the learning data using the second learning program, and generates, for example, second learning result data shown in FIG. The method of machine learning is machine learning (deep learning) using a multilayer neural network (Deep Neural Network) as in the first embodiment described above.

  As described above, the output used to generate the learning data is a specific output, but the output data includes various types of outputs, and includes an output that changes with time. The base station 300A learns not only the instantaneous value of specific output data, but also various types of output included in output data, and output and state data that change with time, when performing machine learning on specific components. I do. Thus, it is possible to generate second learning result data that can determine the relationship between various types of output and state data and the relationship between the output and state data in the transition of time. That is, the base station 300A machine-learns the relationship between various types of output and state data, and the relationship in the transition of time between output and state data, and adjusts the weight and bias of the neural network serving as a model. A neural network for determining the relationship between the output data and the state of the component is generated.

  The state of the component is indicated by, for example, evaluation values 1 to 10 after a predetermined time. The predetermined time is, for example, one hour. Evaluation value 1 has the highest degree of failure, and evaluation value 10 has the lowest degree of failure. That is, the larger the evaluation value, the smaller the degree of the defect. The second learning result data is generated for each component.

  FIG. 12 is a conceptual diagram of a neural network (second learning result data) for determining the state of the battery. In FIG. 12, a plurality of types of outputs are input to the input layer, and are output by the hidden layer (one layer is shown in FIG. 12, but there are actually a plurality of hidden layers) and the output layer. A neural network is shown. The state of the battery is given by the learning data. The base station 300A sets weights and biases so as to minimize or eliminate the difference between the result of processing the input of the power supplied to the motor, the motor sound, and the like by the neural network and the state shown in the learning data. Adjust to generate a second learning result.

  The base station 300A applies the output data newly received at the present time to the second learning result data, that is, the neural network of FIG. The current state of the component such as 1A and the state after a predetermined time are determined. The predetermined time is, for example, one hour.

  The base station 300A notifies the user of the specific unmanned aerial vehicle 1A or the like by the second notification program if a problem occurs in the current component (for example, when the evaluation value is 5 or less), and further, If a malfunction occurs after a predetermined time (for example, when the evaluation value after the predetermined time becomes 5 or less), the user of the specific unmanned aerial vehicle 1A or the like is notified of the fact.

  The base station 300A updates the second learning result data by the reinforcement learning program. As described above, the base station 300A receives the output data at predetermined time intervals from the drone 1A or the like, and updates the learning data. The base station 300A applies the output data received when the learning data is updated to the second learning result data before the update, and determines the state of the component. If there is a difference between the state of the component included in the learning data and the state of the component determined using the second learning result data before update, the base station 300A determines that the second learning result Adjust the bias and weight of the data to minimize the difference.

  Hereinafter, the processing by the base station 300A will be described with reference to FIG. Base station 300A has received output data from drone 1A or the like (step ST101 in FIG. 13), and generates learning data using a specific output included in the output data (step ST102). Then, base station 300A machine-learns the learning data to generate second learning result data (step ST103). Subsequently, base station 300A applies the newly received output data to the second learning result data (step ST104), and determines whether it is time to replace any of the components within a predetermined time (step ST105). ). If base station 300A determines that it is time to replace any part within a predetermined time, it notifies the user of the state of the part (step ST106). The base station 300A repeats steps ST101 to ST106 until it determines that the termination condition is satisfied, such as receiving a notification from a specific user that notification of component status is unnecessary (step ST107).

  When base station 300A determines whether it is time to replace any part within a predetermined time (step ST105), base station 300A determines the state of the part at that time, and determines that the part has a defect. When it is determined, the user is notified of a failure of the part, such as a specific unmanned aerial vehicle 1A, on which the part is mounted.

In addition, the status notification device of the parts of the unmanned aerial vehicle, the method of notifying the status of the components of the unmanned aerial vehicle, and the program according to the present invention are not limited to the above embodiments, and may be variously modified without departing from the gist of the present invention. Can be.

1A-1H Unmanned aerial vehicle (unmanned aerial vehicle)
2 Housing 4 Arm 6 Motor 8 Propeller 10 Protective frame 12 Fixing device 14 Camera 16 Antenna 300 300A Base station

Claims (11)

Learning result data storage means for storing learning result data obtained by machine learning the relationship between a plurality of different use conditions of the unmanned aerial vehicle and the durability of parts of the unmanned aerial vehicle,
Scheduled condition receiving means for receiving scheduled condition data indicating a scheduled use condition of a specific unmanned flight device;
The learning result data, the scheduled condition data, and a state determination unit that determines a state of a component of the specific unmanned aerial vehicle based on a lapse of time,
For the user of the specific unmanned aerial vehicle, a notification transmitting unit that transmits a notification regarding a state of a part of the specific unmanned aerial vehicle,
A status notification device for parts of an unmanned aerial vehicle having: Second learning result data storage means for storing second learning result data obtained by machine learning the output of the parts and the state of the parts of the unmanned aerial vehicle,
Output data receiving means for receiving output data indicating an output of a specific unmanned aerial vehicle component,
Based on the second learning result data and the output data, a second state determination unit that determines a state of a part of the specific unmanned aerial vehicle,
For the user of the specific unmanned aerial vehicle, a second notification transmitting unit that transmits a notification regarding the state of the part of the specific unmanned aerial vehicle,
A status notification device for parts of an unmanned aerial vehicle having:   3. The device status notification device according to claim 2, further comprising a second learning unit configured to machine-learn the output and generate the second learning result data. 4. Based on the specific output, generating state data indicating the state of the component, has a learning data generating means for generating learning data composed of the output data and the state data,
The second learning means is configured to learn the learning data,
The device for notifying the status of a part of an unmanned aerial vehicle according to claim 3. The second state determination means is configured to determine a future state of the component,
The unmanned unmanned aerial vehicle according to claim 3, wherein the second notifying unit is configured to notify a user of the specific unmanned aerial vehicle of a future state of the component. Status notification device for parts of flight equipment.   The second learning means is configured to machine-learn the learning data that changes with the passage of time, and to generate the second learning result data capable of determining a future state of the component. An apparatus for notifying the status of a component of an unmanned aerial vehicle according to any one of claims 3 to 5.   Applying the set of outputs received when generating the learning data to the second result data, the state data included in the learning data, and the component determined by applying the second result data The unmanned flying device according to any one of claims 2 to 6, further comprising a reinforcement learning unit that adjusts the second result data when there is a difference from the state, and cancels the difference. Component status notification device. Status notification of parts of an unmanned aerial vehicle having learning result data storage means for storing learning result data obtained by machine learning the relationship between a plurality of different use conditions of the unmanned aerial vehicle and the durability of the parts of the unmanned aerial vehicle The device
A scheduled condition receiving step of receiving scheduled condition data indicating a scheduled use condition of a specific unmanned flight device;
The learning result data, the scheduled condition data, and a state determining step of determining a state of a part of the specific unmanned aerial vehicle based on the passage of time,
For the user of the specific unmanned aerial vehicle, a notification transmitting step of transmitting a notification regarding the state of the part of the specific unmanned aerial vehicle,
For notifying the status of parts of an unmanned aerial vehicle that performs the operation. An unmanned aerial vehicle component status notification device having second learning result data storage means for storing second learning result data obtained by machine learning the output information of the unmanned aerial vehicle component,
An output data receiving step of receiving output data indicating an output of a specific unmanned aerial vehicle;
Based on the second learning result data and the output data, a second state determination step of predicting the state of the component of the specific unmanned aerial vehicle,
For the user of the specific unmanned aerial vehicle, a second notification transmitting step of transmitting a notification regarding the state of the part of the specific unmanned aerial vehicle,
For notifying the status of parts of an unmanned aerial vehicle that performs the operation. A computer that controls an unmanned aerial vehicle component status notification device,
Learning result data storage means for storing learning result data obtained by machine learning the relationship between a plurality of different use conditions of the unmanned aerial vehicle and the durability of parts of the unmanned aerial vehicle,
Scheduled condition receiving means for receiving scheduled condition data indicating a scheduled use condition of a specific unmanned flight device,
The learning result data, the scheduled condition data, and a state determination unit that determines a state of a component of the specific unmanned aerial vehicle based on a lapse of time,
Notification transmission means for transmitting a notification regarding the status of parts of the specific unmanned aerial vehicle to the user of the specific unmanned aerial vehicle,
Program to function as A computer that controls an unmanned aerial vehicle component status notification device,
Second learning result data storage means for storing second learning result data obtained by machine learning the output information of the parts of the unmanned aerial vehicle,
Environmental condition data receiving means for receiving output data indicating an output of a specific unmanned flight device,
Based on the second learning result data and the output data, a second state determination unit that predicts a state of a part of the specific unmanned aerial vehicle,
For the user of the specific unmanned aerial vehicle, a second notification transmitting unit that transmits a notification regarding the state of the component of the specific unmanned aerial vehicle,
Program to function as