JP7097006B2 - Sensor system - Google Patents

Description

The present invention relates to a sensor system.

Conventionally, in order to improve the accuracy of a device such as a sensor, noise is predicted by a learning model generated by using a machine learning method, and noise is removed from the data output by the device.

For example, Patent Document 1 discloses an expert diagnostic method that automatically replaces a learning model, newly learns and trains, and forms a new determination system. The expert diagnostic method includes steps to automatically replace the learning model and update the expert system.

Further, Patent Document 2 discloses a power consumption estimation device including a learning / classifier 12 as a power consumption estimation device and a data exchange unit 15 for exchanging a learning model with another learning model. The power consumption estimation device determines which learning pattern the network traffic pattern is closest to, and replaces it with an appropriate learning model.

Special Table 2010-591691 Gazette Japanese Unexamined Patent Publication No. 2013-97713

However, the learning model is not always improved by learning. For example, overfitting may occur in which performance deteriorates due to learning. Then, when a learning model having deteriorated performance is applied to a predetermined device such as a sensor, the entire system using the predetermined device is adversely affected. Therefore, it may be necessary to evaluate the learning model.

Therefore, the present invention has been made in view of such circumstances, and an object of the present invention is to provide a sensor system capable of suppressing deterioration of the performance of a learning model.

The sensor system according to one aspect of the present invention is a determination unit that makes a predetermined determination using an acquisition unit that acquires sensing data output by the sensor and an output value obtained by inputting the sensing data into the first learning model. A learning unit that generates a second learning model by machine learning using sensing data, a learning model determination unit that determines which of the performance of the first learning model and the performance of the second learning model is superior, When the learning model determination unit determines that the performance of the second learning model is superior to the performance of the first learning model, it includes an alternative unit that substitutes the first learning model with the second learning model.

According to this aspect, the determination by the sensing data is performed by the first learning model, and the second learning model is further generated by the sensing data. Then, the first learning model is replaced by the second learning model only when it is determined that the performance of the second learning model is superior to the performance of the first learning model. Therefore, deterioration of the learning model due to overfitting or the like is suppressed.

According to the present invention, it becomes possible to provide a sensor system capable of suppressing deterioration of the performance of a learning model.

It is a figure which shows the outline of the sensor system 1 which concerns on 1st Embodiment. It is a figure which shows the functional block of the master unit 10 which concerns on 1st Embodiment. It is a figure for demonstrating the generation process of the learning model by the sensor system 1 which concerns on 1st Embodiment. It is a figure for demonstrating the performance determination process of the learning model by the sensor system 1 which concerns on 1st Embodiment. It is a figure for demonstrating the performance determination process of the learning model by a sensor system 1. It is a schematic diagram for demonstrating the structure of the bite detection part 91 which concerns on 2nd Embodiment. It is a figure which shows the functional block of the master unit 10 which concerns on 2nd Embodiment. It is an operation flow diagram for demonstrating the generation process and the performance determination process of the learning model by the sensor system 1 which concerns on 2nd Embodiment. It is an operation flow diagram for demonstrating the learning process of S207. It is an operation flow diagram for demonstrating the learning model determination process of S303. It is a figure which shows the outline of the sensor system 1 which concerns on 3rd Embodiment. It is a figure which shows the functional block of the master unit 10 which concerns on 3rd Embodiment. It is an operation flow diagram of the generation processing of the learning model by the sensor system 1 which concerns on 3rd Embodiment. This is an example of the learning progress display screen. This is an example of the display screen of the learning completion notification. It is a figure which shows another example of the screen of the learning progress display. It is a figure which shows an example of the system by the connection composition using a plurality of sensors. It is a figure for demonstrating the learning section in the system by the connection composition using a plurality of sensors.

Hereinafter, an embodiment according to one aspect of the present invention (hereinafter, also referred to as “the present embodiment”) will be described with reference to the drawings. In each figure, those having the same reference numerals have the same or similar configurations.

[First Embodiment]
<Sensor system 1>
FIG. 1 is a diagram showing an outline of the sensor system 1 according to the first embodiment. The sensor system 1 includes a master unit 10, a first slave unit 20a, a second slave unit 20b, a third slave unit 20c, a first sensor 30a, a second sensor 30b, a third sensor 30c, and a PLC 40. The first sensor 30a, the second sensor 30b, and the third sensor 30c are arranged along the line L, and measure data indicating the passing state of the work conveyed on the line L. The first sensor 30a, the second sensor 30b, and the third sensor 30c may each be composed of, for example, a fiber sensor. Alternatively, these sensors 30 may be transmission type or regression reflection type photoelectric sensors. Alternatively, these sensors 30 may be displacement sensors that project a laser beam onto the object and obtain a signal value corresponding to the distance to the object based on the principle of triangular distance measurement. Alternatively, these sensors 30 may be distance measuring sensors that obtain a signal value corresponding to the distance to the object based on the round-trip time of the light reflected by the object.

The first slave unit 20a, the second slave unit 20b, and the third slave unit 20c are connected to each of the plurality of sensors 30 and acquire data measured by the plurality of sensors 30. More specifically, the first slave unit 20a is connected to the first sensor 30a, the second slave unit 20b is connected to the second sensor 30b, and the third slave unit 20c is connected to the third sensor 30c.

PLC40 is a control device. The master unit 10 is connected to a plurality of slave units 20a and a control device. In the present specification, the first slave unit 20a, the second slave unit 20b, and the third slave unit 20c are collectively referred to as the slave unit 20, and the first sensor 30a, the second sensor 30b, and the third sensor 30c are collectively referred to as the sensor 30. .. The configuration of the sensor system 1 according to the present embodiment is an example, and the number of the plurality of sensors 30 and the number of the plurality of slave units 20a included in the sensor system 1 are arbitrary. Further, the control device does not necessarily have to be PLC40.

The master unit 10 may be connected to the PLC 40 via a communication network such as a LAN (Local Area Network). The slave unit 20 is physically and electrically connected to the master unit 10. In the present embodiment, the master unit 10 stores the information received from the slave unit 20 in the storage unit, and transmits the stored information to the PLC 40. Therefore, the data acquired by the slave unit 20 is unified by the master unit 10 and transmitted to the PLC 40.

As an example, a determination signal and detection information are transmitted from the slave unit 20 to the master unit 10. The determination signal is a signal indicating a determination result regarding the work, which is determined by the slave unit 20 based on the data measured by the sensor 30. For example, when the sensor 30 is a photoelectric sensor, the determination signal may be an on signal or an off signal obtained by comparing the received light amount measured by the sensor 30 with the threshold value by the slave unit 20. The detection information is a detection value obtained by the detection operation of the slave unit 20. For example, when the sensor 30 is a photoelectric sensor, the detection operation may be the operation of projecting light and receiving light, and the detection information may be the amount of light received.

The slave unit 20 may be attached to the side surface of the master unit 10. Parallel communication or serial communication may be used for communication between the master unit 10 and the slave unit 20. That is, the master unit 10 and the slave unit 20 may be physically connected by a serial transmission line and a parallel transmission line. For example, the determination signal may be transmitted from the slave unit 20 to the master unit 10 on the parallel transmission line, and the detection information may be transmitted from the slave unit 20 to the master unit 10 on the serial transmission line. The master unit 10 and the slave unit 20 may be connected to either a serial transmission line or a parallel transmission line.

<Master unit 10>
FIG. 2 is a diagram showing a functional block of the master unit 10 according to the first embodiment. The master unit 10 is an example of a control device, and includes a learning progress display unit 11, a learning model determination unit 12, a communication unit 14, a determination unit 15, a determination display unit 16, an output unit 17, and an input unit 18. These may be realized by one processor included in the master unit 10, or may be realized by a plurality of processors.

The learning progress display unit 11 is a means for displaying the degree of learning progress of the learning model 2 and the substitution of the learning model 1, and is composed of, for example, an LED or a liquid crystal display.

The learning model determination unit 12 compares the performance of the learning model 1 with the performance of the learning model 2 and determines which performance is better. Any method may be adopted for determining the performance of the learning model. For example, the determination method may be defined as a learning model in which the smaller the variance value of the difference between the sensing data and the predicted data in a predetermined period is, the better the learning model is. Alternatively, the determination method may be defined as a learning model in which the smaller the absolute value of the maximum value and the minimum value of the difference between the sensing data and the predicted data in a predetermined period, the better. Alternatively, the determination method may be defined as a better learning model as the average value of the differences between the sensing data and the predicted data in a predetermined period is smaller.

Further, the learning model determination unit 12 may transmit the determination result of the performance of the learning model to a higher-level unit such as the PLC 40 via the learning progress display unit 11 or the communication unit 14. Further, the learning model determination unit 12 may determine the performance of the learning model at regular intervals, for example.

The learning unit 13 performs machine learning using the sensing data as learning data between sensings, generates a learning model 2, and stores the learning model 2 in the storage unit 20. The communication unit 14 is an interface for communicating with the PLC 40. The communication unit 14 may communicate with a higher-level unit other than the PLC 40. The communication unit 14 may be a communication module capable of wired communication such as a fieldbus, or a communication module capable of wireless communication.

The determination unit 15 uses the learning model 1 to determine the work based on the sensing data. The determination display unit 16 displays the determination result. The mode of display is not particularly limited. The output unit 17 outputs the determination result to an external device such as a PLC 40. The output value may be an On / Off binary value or an encoded multi-bit signal. The input unit 18 is composed of buttons and the like that accept user operations. The input unit 18 may accept, for example, an input for rolling back the learning data.

The sensing unit 19 is composed of a slave unit 20a, a processing unit 19a, and an acquisition unit 19b. The processing unit 19a and the acquisition unit 19b may or may not be included in the master unit 10, for example. The processing unit 19a processes the data received from the slave unit 20a. The mode of processing is not particularly limited. The acquisition unit 19b acquires the data processed by the processing unit 19a and stores it in the storage unit 20 as sensing data.

The storage unit 20 is composed of one or a plurality of memories. The storage unit may have one access bus or a plurality of access buses. The storage unit 20 stores, for example, sensing data, a data processing method, a learning model 1, a learning model 2, an old learning model, a learning determination method, and a model 1 determination result.

The sensing data may be the data processed by the processing unit 19a without processing the data measured by the sensor 30. The data processing method is a method that defines a process for processing sensing data into learning data. The processing method is not particularly limited, and may include a case of "not processing" (data measured by the sensor 30 is used as it is). The learning model 1 (first learning model) is a learning model used for determining a work. The learning model 2 (second learning model) is a learning model generated by the learning unit 13. The old learning model is a learning model related to the learning model 1 before the learning model 1 is replaced by the learning model 2. The learning determination method is data that defines a determination method for determining the performance of the learning model. The model 1 determination result is data showing the determination result regarding the performance of the learning model 1.

The master unit 10 may have a different pair of learning models for each slave unit. In this case, for example, the first slave unit 20a corresponds to the learning model 1 and the learning model 2, and the second slave unit 20b has the learning model 3 (third learning model) and the learning model 4 (fourth learning model). handle.

<Learning model generation process>
FIG. 3 is a diagram for explaining a learning model generation process by the sensor system 1 according to the first embodiment.

First, the learning unit 13 reads the data processing method stored in the storage unit 20. The data processing method is not particularly limited, and may include, for example, averaging processing, FFT (Fast Fourier Transform) processing, smoothing processing, and the like. Next, the learning unit 13 acquires the sensing data stored in the storage unit 20 and processes the sensing data by the above data processing method to generate learning data. Next, the learning unit 13 performs machine learning using the generated learning data to generate a learning model. Next, the learning unit 13 stores the generated learning model as the learning model 2 in the storage unit 20. This completes the learning model generation process.

<Learning model performance judgment processing>
4A and 4B are diagrams for explaining the performance determination process of the learning model by the sensor system 1 according to the first embodiment.

First, the learning model determination unit 12 acquires sensing data from the storage unit 20, and post-processes the sensing data by the post-processing unit 1, the post-processing unit 2, and the like based on the learning determination method. The content of the post-processing is not particularly limited, but may include, for example, an averaging process, an FFT process, a subtraction / addition process, and the like. Next, the learning model determination unit 12 performs prediction processing based on the learning model 2. Next, the learning model determination unit 12 calculates the determination result by the determination result calculation unit. The content of the determination result calculation method is not particularly limited, but for example, as described above, the determination result may be calculated using the variance value, the maximum / minimum value, the average value, and the like. Next, the learning model determination unit 12 compares the determination results by the determination result comparison / reflection unit. At this time, if the determination result that the performance of the learning model 2 is superior to the performance of the learning model 1 is calculated, the learning model 1 is substituted by the learning model 2. This completes the performance determination process of the learning model.

[Second Embodiment]
Hereinafter, the second embodiment will be described. The second embodiment is an example in which the present invention is applied to the bite detection unit 91 of the packaging device. Hereinafter, the portion of the second embodiment that is different from the first embodiment will be described, and the description of the same configuration as that of the first embodiment will be omitted as appropriate.

<Biting detection unit 91>
FIG. 5 is a schematic diagram for explaining the configuration of the bite detection unit 91 according to the second embodiment. The seal portions 61 and 62 are provided with two sets of welding cutting portions, A1 and B1, A2 and B2, and a sampling period is set for each set of these welding cutting portions according to the timing at which the welding cutting process is performed. To. If foreign matter is caught between the heater portions 81a, 81b, 82a, 82b and the cutter portions 83, 84 when the film material 101 is welded, the rotation axes of the sealing portions 61, 62 shift in the direction perpendicular to the axis. Then, since the metal plate 90 is displaced according to the deviation of the rotation axis, the bite can be detected by measuring the displacement with the proximity sensor 91.

FIG. 6 is a diagram showing a functional block of the master unit 10 according to the second embodiment. The master unit 10 according to the second embodiment acquires sensing data from the bite detection unit 91 instead of the plurality of slave units 20a.

<Learning model generation and learning model performance judgment processing>
FIG. 7 is an operation flow diagram for explaining the learning model generation process and the performance determination process by the sensor system 1 according to the second embodiment.

First, the data processing method is uploaded to the master unit 10 by the upper unit (S201). The data processing method is not particularly limited, but for example, "data is not processed, and n-dimensional data of tn, nt + 1, ..., T (t: measurement time, n: integer) is input to the learning model. You may upload the setting "to" as a data processing method. Next, the learning determination method is uploaded to the master unit 10 by the upper unit (S202). The learning determination method is not particularly limited, but for example, the smaller the “dispersion value Σ | f (t) −p | 2 (f (t): predicted value by the learning model at time t, p: measured value) for a certain period of time). You may upload the setting "the performance of the learning model is excellent" as the learning judgment method.

Next, the model 1 determination result is set to the worst value "0%" by the upper unit (S203). Next, the operation of the bite detection unit 91 is started (S204). Next, the learning of the master unit 10 is started by the upper unit (S205). Next, the end flag is set to 0 and the time is set to 1, respectively (S206). Next, the learning process is executed (S206). The execution of the learning process (S206) is repeated until the end flag becomes 1 (S208).

FIG. 8 is an operation flow diagram for explaining the learning process of S207. First, it is determined whether or not the period specified as the predetermined learning interval has elapsed (S301). Specifically, it is determined whether or not the remainder obtained by dividing the time by the learning interval is 0, and this is repeated until the determination result is Yes. When the determination result of S301 is Yes, the learning model 2 is generated / substituted according to the data processing method stored in the storage unit 20. Next, the learning model determination process is performed (S303).

FIG. 9 is an operation flow diagram for explaining the learning model determination process of S303. First, it is determined whether or not a period specified as a predetermined determination interval has elapsed (S401). Specifically, it is determined whether or not the remainder obtained by dividing the time by the determination interval is 0, and this is repeated until the determination result is Yes. When the determination result of S401 is Yes, the performance of the learning model 2 is evaluated according to the learning determination method stored in the storage unit 20. Next, it is determined whether or not the performance of the learning model 2 is superior to the performance of the learning model 1 (S403).
If the determination result is No, the process ends. On the other hand, when the determination result is Yes, the learning model 2 is copied to the learning model 1 stored in the storage unit 20 (S404). Next, the result of the learning model 2 is stored in the model 1 determination result (S405). This completes the learning model generation process and performance determination process.

[Third Embodiment]
<Sensor system 1>
FIG. 10 is a diagram showing an outline of the sensor system 1 according to the third embodiment. Hereinafter, the parts of the third embodiment that are different from those of the first embodiment will be described, and the parts common to both embodiments will be omitted as appropriate. The sensor system 1 includes a master unit 10, a first slave unit 20a, a second slave unit 20b, a third slave unit 20c, a first sensor 30a, a second sensor 30b, a third sensor 30c, a PLC 40, and an external storage medium 50. .. The external storage medium 50 may be, for example, a PC, a USB memory, or the like. The sensor system 1 can utilize the previous learning model and sample data stored in the external storage medium 50. Further, the sensor system 1 can copy and use the data acquired by another system. The master unit 10 includes a display unit 60. The display unit 60 displays, for example, the progress of the current learning and the learning completion determination. As a result, the user can confirm the progress of the current learning and the learning completion determination.

<Master unit 10>
FIG. 11 is a diagram showing a functional block of the master unit 10 according to the third embodiment. The master unit 10 is an example of a control device, and includes a communication unit 14, a determination unit 15, an acquisition unit 19b, a correction unit 19c, a timer 21, and a display unit 60. These may be realized by one processor included in the master unit 10, or may be realized by a plurality of processors. Further, the storage unit 20 included in the master unit 10 stores, for example, sensing data, measurement timing, and a trained model.

FIG. 12 is an operation flow diagram of a learning model generation process and the like by the sensor system 1 according to the third embodiment. First, the master unit 10 executes a predetermined setting process related to the start of learning (S501). Next, the master unit 10 accepts the input of sample data (S502). Next, the master unit 10 displays the learning progress display screen via the display unit 60 (S503). Next, when the learning is completed, the master unit 10 displays the learning completion notification screen via the display unit 60 (S504).

FIG. 13 is an example of a screen for displaying the learning progress. The learning progress display screen is a display screen showing the ratio of the current model determination accuracy to the target accuracy (target accuracy). The user can grasp the learning progress status from the learning progress display screen. The sensor system 1 displays the determination accuracy of the model and the learning status (progress) by inputting sample data. For the determination accuracy of the model, the correct answer rate for the test data for the determination rate analysis may be used. Specifically, it may be a cross-validation result (hold-out method, leave-one-out method, etc.) for the input sample data, or a correct answer rate for specific data specified by the user. Further, the correct answer rate (target accuracy) of the target learning model may be arbitrarily set by the user. In addition, the number of samples required for prediction may be calculated by obtaining a regression curve from the number of past training samples and the transition of model determination accuracy.

FIG. 14 is an example of a display screen of the learning completion notification. The display screen of the learning completion notification is a display screen showing that the accuracy of the model determination has reached the target accuracy. After the training is completed, it may be possible to select whether to update or cancel the training model. Specifically, as shown in FIG. 14, a selection unit for selecting learning model update and cancellation is displayed on the display screen of the learning completion notification. When training model update is selected, the old training model is updated by the model that has completed training. When Cancel is selected, the learning model update is cancelled. Further, the sensor system 1 may compare the determination accuracy of the past learning model with the determination accuracy of the current learning model and automatically update when the performance is improved. Further, in the sensor system 1, the learning model may be backed up every time the model is updated, and may be restored to the backed up learning model at an arbitrary timing.

FIG. 15 is a diagram showing another example of the learning progress display screen. As shown in FIG. 15, the learning progress display screen may include each numerical value of target discrimination accuracy, model discrimination accuracy, learning progress rate, and a display bar showing the learning progress rate.

<Learning status display with connected amplifier>
FIG. 16 is a diagram showing an example of a system having a connected configuration using a plurality of sensors. In this system, the output signals of some slave units are used as trigger input sensors, and the output signals of some other slave units are used as data acquisition sensors. Specifically, the system has a plurality of slave units 20. Each slave unit 20 has a data acquisition sensor 1 for learning, a trigger sensor 1 for designating a start, a trigger sensor 2 for designating an end, a data acquisition sensor 2 for learning, and a trigger sensor for designating a start. Corresponds to 3 and the trigger sensor 4 that specifies the end. Each slave unit 20 has an indicator light. The indicator light can indicate a state such as learning, triggering, learning completion, and AI invalidation by the display mode shown in FIG. 16, for example. With this configuration, by individually displaying the sensor status by the display lamp color or the like, it is possible to grasp at a glance the role (state) of each sensor and whether or not learning is completed. The sensor status may be displayed on the slave unit in the color of the indicator lamp, or may be displayed in a form in which each status can be identified by the blinking pattern of the indicator lamp, the 7-segment display, or the dot matrix display. An indicator light may be provided in the master unit instead of the slave unit or together with the slave unit.

FIG. 17 is a diagram for explaining a learning section in a system based on a connected configuration using a plurality of sensors. As shown in FIG. 17, the data acquisition sensor 1 starts learning when the signal of the trigger sensor 1 becomes High, and ends learning when the signal of the trigger sensor 2 becomes High. Similarly, the data acquisition sensor 2 starts learning when the signal of the trigger sensor 3 becomes High, and ends learning when the signal of the trigger sensor 4 becomes High. In this way, when the slave unit is used as a trigger sensor, the learning section of the signal can be specified by designating the data acquisition sensor associated with the signal.

As described above, the above-described embodiment has a particularly advantageous effect in the following two cases.
(1) Initial learning At the time of the first learning, the learning model is updated by using the sensing data as learning data at any time. Therefore, in the conventional technique, the performance of the learning model may deteriorate due to overfitting. However, in the above-described embodiment, only the learning model 2 is updated (updated) at any time, and the learning model 1 is updated only when the performance of the learning model 2 is determined to be "better" than the performance of the learning model 1. .. Therefore, the learning model 1 of the master unit 10 always gives the best result.

(2) Real-time learning (unsupervised learning)
Since there are two learning models, it is possible to update the learning model at any time (real-time learning) while operating it as a sensor. Since the cutter part of the packaging device wears and its state changes with the passage of time, the measurement waveform that can be acquired by the proximity sensor also changes. However, this change is not abrupt but gradual. At this time, the learning model is updated with the data predicted by the learning model 1 and the data obtained when the variance value of the measured value for a certain period of time is within the model 1 determination result, that is, when it is determined that the data is not biting. Let me. As a result, the learning model 2 is updated to a learning model that can respond to changes in the environment such as the cutter portion while measuring. Further, since the learning model 1 is updated by determining the quality of the learning model by the learning determination method, it does not become worse. Therefore, since there are two learning models, it is possible to evolve into a learning model that responds to changes in the environment while measuring.

1 Sensor system 10 Master unit 11 Learning progress display unit 12 Learning model judgment unit 13 Learning unit 14 Communication unit 15 Judgment unit 16 Judgment display unit 17 Output unit 18 Input unit 19 Sensing unit 19a Processing unit 19b Acquisition unit 20 Storage unit 20a First Slave unit 20b 2nd slave unit 20c 3rd slave unit 30 Sensor 30a 1st sensor 30b 2nd sensor 30c 3rd sensor 40 PLC
91 Biting detector

Claims (9)

An acquisition unit that acquires sensing data output by a predetermined sensor,
A determination unit that makes a predetermined determination using an output value obtained by inputting the sensing data into the first learning model, and a determination unit.
A learning unit that generates a second learning model by machine learning using the sensing data,
A learning model determination unit that determines which of the performance of the first learning model and the performance of the second learning model is superior,
When the learning model determination unit determines that the performance of the second learning model is superior to the performance of the first learning model, the alternative unit that substitutes the first learning model by the second learning model and the alternative unit.
Equipped with
The learning model determination unit uses the sensing data and the sensing data as the first difference between the sensing data and the output value obtained by inputting the sensing data into the first learning model, and the sensing data and the sensing data. 2 It is determined whether the performance of the first learning model or the performance of the second learning model is superior based on the second difference from the output value obtained by inputting to the learning model. Characteristic control device. The learning model determination unit uses at least one of the variance value, the maximum value, the minimum value, and the average value of the first difference and the second difference in a predetermined period, and uses the first learning model. The control device according to claim 1 , wherein it is determined which of the performance of the second learning model and the performance of the second learning model is superior. The control device according to claim 1 or 2 , wherein the predetermined sensor is a proximity sensor or a bite detection unit. The control device according to any one of claims 1 to 3 , further comprising a processing unit for processing the sensing data output by the predetermined sensor. The control device according to any one of claims 1 to 4 , further comprising a learning progress display unit that displays the progress of determination by the learning model determination unit. The control device according to any one of claims 1 to 5 , further comprising a display unit for displaying the result of determination by the learning model determination unit. The control device according to claim 6 , wherein the display unit displays the result of the determination in the mode according to the first evaluation method and / or in the mode according to the second evaluation method. Further equipped with an operation unit for generating alternative instructions for the learning model,
The alternative unit is a case where the learning model determination unit determines that the performance of the second learning model is superior to the performance of the first learning model, and the instruction is generated by the operation unit. The control device according to any one of claims 1 to 7 , wherein the second learning model replaces the first learning model. The acquisition unit further acquires the second sensing data output by the second sensor, and obtains the second sensing data.
The determination unit further makes another predetermined determination using the output value obtained by inputting the second sensing data into the third learning model.
The learning unit further generates a fourth learning model by machine learning using the second sensing data.
The learning model determination unit further determines which of the performance of the third learning model and the performance of the fourth learning model is superior.
The alternative unit further substitutes the third learning model with the fourth learning model when the learning model determination unit determines that the performance of the fourth learning model is superior to the performance of the third learning model. ,
The control device according to claim 1.

Images