Control system

Abstract

To provide a control system capable of optimizing arithmetic processing by using an inference model.SOLUTION: A control system 1 includes a first intelligent device 10 having a first inference device 13 for performing arithmetic processing of processing information by using a first inference model, and a distributed processing setting part 30. Also, a communication part can communicate with one or more second intelligent devices 20 other than the first intelligent device 10. Then, the distributed processing setting part 30 extracts a candidate for the second intelligent device 20 having a second inference device 22 for performing arithmetic processing using a second inference model from the one or more second intelligent devices 20, and determines validity of making the second inference device 22 in the second intelligent device 20 of the extracted candidate perform arithmetic processing of second partial information being remaining processing information by using the second inference model.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to control systems. [Background technology]

[0002] Patent Document 1 discloses computing that presents a distributed array of candidate resources, relationships, and experience opportunities that are unimaginable from a human perspective. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2018-92632 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in recent years, there has been a demand for further optimization of computational processing using inference models.

[0005] Therefore, an object of the present disclosure is to provide a control system capable of optimizing calculation processing using an inference model. [Means for solving the problem]

[0006] A control system according to one embodiment of the present disclosure includes a first inference device that processes processing information using a first inference model, a first intelligent device having a communication unit, and a distributed processing setting unit, wherein the first inference device processes first partial information, which is a portion of the processing information, using the first inference model, and the communication unit is capable of communicating with one or more second intelligent devices that are devices other than the first intelligent device, and the distributed processing setting unit extracts, from among the one or more second intelligent devices, a candidate for the second intelligent device that has a second inference device that performs arithmetic processing using the second inference model, and determines the appropriateness of having the second inference device in the extracted candidate second intelligent device process second partial information, which is the remaining processing information of the processing information, using the second inference model. Effect of the Invention

[0007] The control system of the present disclosure is capable of optimizing computational processes using inference models. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram showing a first control system according to an embodiment. [Diagram 2] FIG. 2 is a block diagram showing a first intelligent device and a second intelligent device. [Diagram 3] FIG. 3 is a flowchart showing the processing operation of the first control system according to the embodiment. [Figure 4] FIG. 4 is a block diagram showing a first intelligent device according to another modified example. [Diagram 5] FIG. 5 is a block diagram showing a device group according to another modified example. [Figure 6] FIG. 6 is a block diagram showing a server, an edge computer, a first intelligent device, and a second intelligent device according to another modified example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. Note that each of the embodiments described below shows a specific example of the present disclosure. Therefore, the numerical values, shapes, materials, components, arrangement and connection of the components, steps, order of steps, and the like shown in the following embodiments are merely examples and are not intended to limit the present disclosure.

[0010] Each figure is a schematic diagram and is not necessarily precisely illustrated. Therefore, for example, the scales and the like do not necessarily match in each figure. Furthermore, in each figure, the same reference numerals are given to substantially the same configurations, and duplicated explanations are omitted or simplified.

[0011] (Embodiment) <Function and configuration> Hereinafter, the function and configuration of a spatial system including a control system 1 according to the embodiment will be described with reference to FIGS.

[0012] FIG. 1 is a block diagram showing a control system 1 according to an embodiment. FIG. 2 is a block diagram showing a first intelligent device 10 and a second intelligent device 20. FIG. 2 illustrates five second intelligent devices 20a, 20b, 20c, 20d, and 20e. The second intelligent devices 20a, 20b, 20c, 20d, and 20e may be collectively referred to as the second intelligent devices 20. The second intelligent devices 20 illustrated in FIG. 2 are merely an example, and the number of the second intelligent devices 20 is not limited to five. The number of the second intelligent devices 20 may be four or less, or six or more.

[0013] As shown in Fig. 1, in the spatial system, arithmetic processing is performed using a plurality of control systems installed in the space of a facility. The facility is, for example, a house, an office building, a store, etc. In this embodiment, a control system 1 and one or more second intelligent devices 20 are used as the plurality of control systems for explanation.

[0014] As shown in FIG. 1, the control system 1 includes a first intelligent device 10 and a distributed processing setting unit 30.

[0015] The first intelligent device 10 includes at least one of a first sensor 11 and a first actuator 12, a first inference device 13, and a first communication unit 14. Note that FIG. 1 illustrates an example in which the first intelligent device 10 includes both the first sensor 11 and the first actuator 12.

[0016] The first sensor 11 is a device installed in a space and capable of detecting the state of the space. The first sensor 11 acquires processing information indicating the activities of people in the space, such as images in the space and sounds in the space. The first sensor 11 also acquires processing information indicating the state of the space, such as temperature and illuminance in the space. The first sensor 11 is, for example, a camera, a microphone, a temperature sensor, an illuminance sensor, etc. The first sensor 11 outputs the acquired processing information to the first inference device 13. The first sensor 11 is included in the sensor of the claims of this application.

[0017] The first actuator 12 is a device that is installed in a space and changes the state of the space by changing brightness, airflow, sound, video, etc. The first actuator 12 can change the state of the space based on the result of calculation processing in the control system 1. The first actuator 12 is, for example, a lighting fixture, an air conditioner, a speaker, a display device, etc. The first actuator 12 is included in the actuator of the claims of this application.

[0018] For example, when the first intelligent device 10 does not have the first sensor 11, the first intelligent device 10 controls the first actuator 12 by acquiring processing information from an external first sensor. For example, when the first intelligent device 10 does not have the first actuator 12, the first intelligent device 10 drives the external first actuator by transmitting control information based on the result of arithmetic processing of the processing information to the external first actuator.

[0019] The first inference device 13 acquires the processing information output by the first sensor 11. The first inference device 13 performs arithmetic processing using a first inference model based on the acquired processing information. The first inference device 13 controls the first actuator 12 based on the result of the arithmetic processing. In other words, the first inference device 13 controls the first actuator 12 according to the result of the arithmetic processing based on the processing information output by the first sensor 11. Here, the first inference model is a neural network including an input layer, an output layer, and multiple intermediate layers. The first inference model is stored in a storage medium such as a memory mounted in the first inference device 13.

[0020] The first inference device 13 is hardware for computational processing including, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a TPU (Tensor Processing Unit), an NPU (Neural network Processing Unit), an ASIC (Application Specific Integrated Circuit), a DSP (digital signal processor), or an Accelerator (Accelerator), etc.

[0021] Furthermore, the first inference device 13 uses the first inference model to perform arithmetic processing of the processing information. However, in this embodiment, when the processing capacity of the first inference device 13 is low or the amount of information is large, if the first inference device 13 performs arithmetic processing of the processing information alone, the arithmetic processing may take a long time. In this case, the first inference device 13 performs arithmetic processing of first partial information, which is a part of the processing information, using the first inference model.

[0022] Here, the processing information is control information for the control system 1 to execute the first actuator 12. For example, if the first actuator 12 is a lighting fixture, the control information is information for turning on and off the lighting fixture, adjusting the brightness and adjusting the color. For example, if the first actuator 12 is an air conditioner, the control information is information for executing the air volume, hot air, cold air, and blowing air of the air conditioner. For example, if the first actuator 12 is a speaker, the control information is information for executing the speaker volume, sound pressure, playing music data, stopping music data, and the like. For example, if the first actuator 12 is a display device, the control information is information for executing the brightness, color tone, playing images, stopping images, and the like of an image displayed by the display device.

[0023] 1 and 2, the first communication unit 14 is a communication interface that enables communication with one or more second intelligent devices 20 that are devices other than the first intelligent device 10. Specifically, the first communication unit 14 is capable of communication with one or more second intelligent devices 20 that are devices other than the first intelligent device 10. The first communication unit 14 is an interface such as Bluetooth (registered trademark), BLE (Bluetooth Low Energy), a wireless LAN (Local area network), or a wired LAN. The first communication unit 14 is included in the communication unit in the claims of the present application.

[0024] The second intelligent device 20 may have the same configuration as the first intelligent device 10. That is, the second intelligent device 20 may have at least one of the second sensor 21 and the second actuator 21e, the second inference device 22, and the second communication unit 23. Therefore, the second sensor 21 may be configured by replacing the first sensor 11 of the present disclosure. Also, the second actuator 21e may be configured by replacing the first actuator 12 of the present disclosure. Also, the second inference device 22 may be configured by replacing the first inference device 13 of the present disclosure. Also, the second communication units 23a, 23b, 23c, 23d, and 23e may be configured by replacing the first communication unit 14 of the present disclosure. Note that the second sensors 21a, 21b, 21c, and 21d may be collectively referred to as the second sensor 21. Also, the second inference devices 22a, 22b, 22c, 22d, and 22e may be collectively referred to as the second inference device 22.

[0025] The second sensor 21a, second inference device 22a, and second communication unit 23a possessed by the second intelligent device 20a in Figure 2, the second sensor 21b, second inference device 22b, and second communication unit 23b possessed by the second intelligent device 20b, the second sensor 21c, second inference device 22c, and second communication unit 23c possessed by the second intelligent device 20c, the second sensor 21d, second inference device 22d, and second communication unit 23d possessed by the second intelligent device 20d, and the second actuator 21e, second inference device 22e, and second communication unit 23e possessed by the second intelligent device 20e may each have different functions.

[0026] The second intelligent device 20 may further include a distributed processing setting unit 30. Moreover, the second intelligent device 20 and the distributed processing setting unit 30 may constitute a control system.

[0027] The distributed processing setting unit 30 of the control system 1 extracts the second intelligent devices 20 that are candidates, and determines the validity of the second inference apparatus 22 in the second intelligent devices 20 that are candidates.

[0028] First, the distributed processing setting unit 30 extracts candidates for the second intelligent device 20 having the second inference apparatus 22 that performs arithmetic processing using the second inference model from among one or more second intelligent devices 20. Specifically, the distributed processing setting unit 30 extracts candidates for the second intelligent device 20 that is one or more second intelligent devices 20 that can communicate with the first intelligent device 10 of the control system 1 and has the second inference apparatus 22 that is suitable for arithmetic processing the second partial information. Here, the second inference model is a neural network including an input layer, an output layer, and multiple intermediate layers.

[0029] For example, the distributed processing setting unit 30 may extract a second intelligent device 20 suitable for arithmetic processing of the second partial information from a database that aggregates information on the functions of one or more second intelligent devices 20. The database may be stored in a storage medium such as a memory mounted in the control system 1, or may be stored in a storage medium of a local server that manages the spatial system. In another example, the distributed processing setting unit 30 may extract a second intelligent device 20 suitable for arithmetic processing of the second partial information by acquiring information on the functions from each of the one or more second intelligent devices 20.

[0030] Here, the information on the functions is the functions that the second intelligent device 20 can actually execute, such as information on the functions of lighting, air conditioning, sound, and image display.

[0031] Next, the distributed processing setting unit 30 can determine the appropriateness of having the second inference device 22 in the extracted candidate second intelligent device 20 perform arithmetic processing of the second partial information, which is the remaining processing information of the processing information, using the second inference model. For example, the distributed processing setting unit 30 can determine the appropriateness of having the second inference device 22 of the second intelligent device 20 perform arithmetic processing of the second partial information using the second inference model, based on information on the functions of one or more second intelligent devices 20.

[0032] The function of the distributed processing setting unit 30 will now be described in more detail.

[0033] The distributed processing setting unit 30 includes a device information collecting unit 31, a host device processing estimating unit 32, a candidate device processing estimating unit 33, and a distributed configuration arrangement unit .

[0034] The device information collection unit 31 collects one or more second intelligent devices 20 with which communication is possible.

[0035] The own device processing estimation unit 32 estimates the processing capability of the first intelligent device 10 and outputs a first result including the estimated processing capability. For example, the own device processing estimation unit 32 estimates the processing capability based on the performance and functions of the first intelligent device 10, which is its own device, and outputs the first estimated result to the distributed configuration arrangement unit 34.

[0036] The candidate device processing estimation unit 33 extracts candidates for the second intelligent device 20 having the second inference device 22 that performs arithmetic processing using the second inference model from among one or more second intelligent devices 20 collected by the device information collection unit 31. As described above, the candidate device processing estimation unit 33 may extract candidates for the second intelligent device 20 using a database, or may extract candidates for the second intelligent device 20 based on information on functions acquired from each of the one or more second intelligent devices 20. Then, the candidate device processing estimation unit 33 estimates the processing capabilities of the second intelligent device 20 that is a candidate from among the one or more collected second intelligent devices 20, and generates one or more second results including the estimated processing capabilities. In addition, when there are multiple candidates, the candidate device processing estimation unit 33 estimates the processing capabilities of the multiple second intelligent devices 20, and generates one or more second results including the estimated processing capabilities. For example, the candidate device processing estimation unit 33 estimates the processing capacity based on the performance and functions of the candidate second intelligent device 20, and outputs one or more second estimated results to the distributed configuration arrangement unit .

[0037] Further, the candidate device processing estimation unit 33 judges the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information based on the first result and one or more second results. For example, when the processing capacity of the second intelligent device 20 shown in the second result is higher than the processing capacity of the first intelligent device 10 shown in the first result and it is judged that the second partial information can be processed, the candidate device processing estimation unit 33 judges that it is appropriate to have the second inference device 22 perform the arithmetic processing of the second partial information. On the other hand, when the processing capacity of the second intelligent device 20 shown in the second result is lower than the processing capacity of the first intelligent device 10 shown in the first result and it is judged that the second partial information cannot be processed, the candidate device processing estimation unit 33 judges that it is not appropriate to have the second inference device 22 perform the arithmetic processing of the second partial information.

[0038] Further, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model based on the spare processing capacity of the second inference device 22. For example, even if the processing capacity of the second inference device 22 is high, the second inference device 22 may not have spare capacity at present because it is performing arithmetic processing of other information. For this reason, the candidate device processing estimation unit 33 causes the second inference device 22, which has spare processing capacity, to perform the arithmetic processing of the second partial information using the second inference model. The spare processing capacity indicates that the second inference device 22 still has spare computing resources and spare processing capacity for the currently executing task and the currently executing process, and is a state in which there is spare capacity for additional tasks and additional arithmetic processing.

[0039] Further, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model based on the similarity between the function of the first inference device 13 and the function of the second inference device 22. For example, when the function of the first inference device 13 and the function of the second inference device 22 are similar, it is considered that the second inference device 22 is suitable for performing the arithmetic processing of the second partial information. Therefore, if the similarity between the function of the first inference device 13 and the function of the second inference device 22 is equal to or higher than a certain level, the candidate device processing estimation unit 33 causes the second inference device 22 to perform the arithmetic processing of the second partial information using the second inference model. Note that, even when the first inference model used in the first inference device 13 and the second inference model used in the second inference device 22 are similar, the candidate device processing estimation unit 33 may cause the second inference device 22 to perform the arithmetic processing of the second partial information using the second inference model.

[0040] In addition, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model based on the type of accelerator constituting the second inference device 22 using the second inference model and the type of operator constituting the second inference model. For example, there are tasks or arithmetic processing that are suitable for a type of accelerator depending on the type of operator. In other words, the optimal accelerator may differ depending on the type of operator in each layer of the neural network. For this reason, the candidate device processing estimation unit 33 may cause the second inference device 22 using the second inference model having a type of operator and a type of accelerator suitable for arithmetic processing of the second partial information to perform arithmetic processing.

[0041] Here, the accelerator is a processing device such as hardware or software, for example, a GPU, an ASIC, a DSP, an image processing Acc, a language processing Acc, etc. Also, the operator refers to a specific type of calculation and operation, for example, a matrix multiplication, a convolution operation, application of an activation function, etc.

[0042] Further, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model based on the type of accelerator constituting the second inference device 22 using the second inference model and the size of the first inference model. Here, the size of the first inference model indicates parameters such as weights and biases included in the convolutional layer constituting the second inference model, the size of the architecture structure of the model, and the like. For example, when the size of the first inference model is large, the arithmetic processing speed of the second inference device 22 may decrease depending on the type of accelerator. For this reason, the candidate device processing estimation unit 33 may cause the second inference device 22 using the second inference model having a type of accelerator suitable for arithmetic processing of the second partial information to perform the arithmetic processing.

[0043] Further, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model based on whether or not the communication speed between the first intelligent device 10 and the second intelligent device 20 is faster than a reference value. For example, when the communication speed is faster than a reference value, a delay is unlikely to occur in the communication between the first intelligent device 10 and the second intelligent device 20. Therefore, the second partial information can be transmitted from the first intelligent device 10 to the second intelligent device 20 at high speed. Therefore, the candidate device processing estimation unit 33 may have the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model when the communication speed is faster than a reference value.

[0044] Further, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model based on the inference speed of one or more second intelligent devices 20. For example, when the inference speed of the second inference device 22 included in the second result is slower than the inference speed of the first inference device 13 included in the first result and is slower than the reference speed, a discrepancy may occur between the inference speed of the first inference device 13 and the inference speed of the second inference device 22. In this case, if the second inference device 22 is caused to perform the arithmetic processing of the second partial information using the second inference model, a delay will occur in the arithmetic processing of the second inference device 22. For this reason, the candidate device processing estimation unit 33 may cause the second inference device 22 to perform the arithmetic processing of the second partial information using the second inference model when the inference speed is faster than the reference speed.

[0045] Furthermore, the candidate device processing estimation unit 33 can determine the appropriateness of having the second inference device 22 perform calculation processing on the second partial information using the second inference model based on the energy consumption of one or more second intelligent devices 20. For example, if a second intelligent device 20 with high energy consumption is selected, the load on the system in which the second intelligent device 20 is mounted may become large. For this reason, the candidate device processing estimation unit 33 may select a second intelligent device 20 with low energy consumption among the second intelligent devices 20 and have the second inference device 22 perform calculation processing on the second partial information.

[0046] As described above, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform the arithmetic processing. As described above, the candidate device processing estimation unit 33 may determine the appropriateness of having the second inference device 22 perform the arithmetic processing based on at least one of the spare processing capacity, the similarity of functions, the relationship between the type of accelerator and the type of operator, the relationship between the type of accelerator and the size of the first inference model, the communication speed, the inference speed, the energy consumption, and the like.

[0047] When the candidate device processing estimation unit 33 determines that it is appropriate to have the second inference device 22 perform arithmetic processing on the second partial information, the distributed configuration arrangement unit 34 causes the second inference device 22 to perform arithmetic processing on the second partial information using the second inference model. In other words, when it is determined that it is appropriate to have the second inference device 22 perform arithmetic processing on the second partial information using the second inference model, the distributed configuration arrangement unit 34 distributes the second partial information to one or more second inference devices 22.

[0048] As a result, the first inference device 13 performs a calculation process on the first partial information using the first inference model, distributes the second partial information to one or more valid second intelligent devices 20 to expand the calculation process, and the second inference devices 22 of the one or more second intelligent devices 20 perform a calculation process on the second partial information using the second inference model. The first intelligent device 10 obtains the result of the calculation process by the first inference device 13 and the result of the calculation process by the second inference devices 22 of the one or more second intelligent devices 20. Based on these results of the calculation processes by the first inference device 13 and the one or more second inference devices 22, the control system 1 can change the state of the space by controlling the first actuator 12.

[0049] <Processing Operation> Next, the processing operation of the control system 1 according to the embodiment will be described with reference to FIGS.

[0050] FIG. 3 is a flowchart showing the processing operation of the control system 1 according to the embodiment.

[0051] First, when the first intelligent device 10 of the control system 1 cannot process all of the processing information to be calculated, or when the calculation takes a long time, the first intelligent device 10 may calculate first partial information, which is a part of the processing information, and have one or more second intelligent devices 20 calculate the remaining second partial information. In this case, the device information collection unit 31 of the distributed processing setting unit 30 in the control system 1 collects one or more second intelligent devices 20 with which it can communicate (S11).

[0052] For example, as shown in FIG. 2, the device information collector 31 can collect information on the existence of a plurality of second intelligent devices 20 that can communicate with the first intelligent device 10. In the example shown in FIG.

[0053] Next, the own device processing estimation unit 32 estimates the processing capability of the first intelligent device 10, and outputs a first result including the estimated processing capability (S12).

[0054] Next, the candidate device processing estimation unit 33 extracts candidates for a second intelligent device 20 having a second inference device 22 that performs calculation processing using a second inference model from among one or more second intelligent devices 20 collected by the device information collection unit 31 (S13).

[0055] 2, for example, when the first intelligent device 10 is mounted on an image sensor, the candidate device processing estimation unit 33 extracts the second intelligent devices 20a and 20b as candidates. As described above, the candidate device processing estimation unit 33 may extract the second intelligent device 20 as a candidate based on at least one of the spare capacity of processing capacity, similarity of functions, relationship between the type of accelerator and the type of operator, relationship between the type of accelerator and the size of the first inference model, communication speed, inference speed, energy consumption, etc. For this reason, the candidate device processing estimation unit 33 may extract the second intelligent devices 20c, 20d, and 20e as candidates.

[0056] Next, the candidate device processing estimation unit 33 estimates the processing capability of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20, and generates one or more second results including the estimated processing capability (S14).

[0057] Next, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing on the second partial information, based on the first result and one or more second results (S15).

[0058] If the candidate device processing estimation unit 33 determines that it is not appropriate to have the second inference device 22 perform arithmetic processing on the second partial information (NO in S15), the distributed configuration arrangement unit 34 terminates the processing operation of Figure 3.

[0059] On the other hand, when the candidate device processing estimation unit 33 judges that it is appropriate to have the second inference device 22 perform the arithmetic processing of the second partial information (YES in S15), the distributed configuration arrangement unit 34 has the second inference device 22 perform the arithmetic processing of the second partial information using the second inference model (S16). In this case, the first inference device 13 performs the arithmetic processing of the first partial information using the first inference model. In this way, the control system 1 has the first intelligent device 10 perform the arithmetic processing of the first partial information, and has the second intelligent device 20 perform the arithmetic processing so as to distribute and expand the second partial information to the appropriate second intelligent device 20. Then, the distributed configuration arrangement unit 34 ends the processing operation of FIG. 3.

[0060] For example, as shown in FIG. 2, the candidate device processing estimation unit 33 estimates the processing capabilities of the candidate second intelligent devices 20a, 20b extracted in step S13.

[0061] The candidate device processing estimation unit 33 determines that the processing capability of the second intelligent device 20a is higher than that of the second intelligent device 20b, and therefore determines that it is appropriate to have the second inference device 22 of the second intelligent device 20a perform the arithmetic processing of the second partial information. As a result, the distributed configuration arrangement unit 34 causes the second inference device 22 of the second intelligent device 20a to perform the arithmetic processing of the second partial information using the second inference model, and outputs the result. Also, the first inference device 13 performs the arithmetic processing of the first partial information using the first inference model, and outputs the result. Based on the results of the arithmetic processing of the first inference device 13 and the second inference device 22, the control system 1 can change the state of the space by controlling the first actuator 12.

[0062] Even if the second intelligent device 20a is appropriate from the viewpoint of processing power, the candidate device processing estimation unit 33 may determine that the second intelligent device 20b is appropriate based on at least one of the spare processing power, similarity of functions, relationship between the type of accelerator and the type of operator, relationship between the type of accelerator and the size of the first inference model, communication speed, inference speed, energy consumption, etc. Furthermore, if the second intelligent devices 20c, 20d, and 20e are further extracted as candidates, the candidate device processing estimation unit 33 may determine that one or more of the second intelligent devices 20a, 20b, 20c, 20d, and 20e are appropriate.

[0063] <Action and effect> Next, the effects of the control system 1 in this embodiment will be described.

[0064] As described above, the control system 1 according to the present embodiment includes the first inference device 13 that performs arithmetic processing of processing information using the first inference model, the first intelligent device 10 having a communication unit (first communication unit 14), and the distributed processing setting unit 30. The first inference device 13 performs arithmetic processing of first partial information, which is a part of the processing information, using the first inference model. The communication unit (first communication unit 14) is capable of communicating with one or more second intelligent devices 20 that are devices other than the first intelligent device 10. The distributed processing setting unit 30 extracts, from among the one or more second intelligent devices 20, a candidate for the second intelligent device 20 that has the second inference device 22 that performs arithmetic processing using the second inference model, and determines whether it is appropriate to have the second inference device 22 in the extracted candidate second intelligent device 20 perform arithmetic processing of the second partial information, which is the remaining processing information of the processing information, using the second inference model.

[0065] According to this, it is possible to extract candidates for the second intelligent device 20 that performs the arithmetic processing from among one or more second intelligent devices 20. Therefore, it is expected that the arithmetic processing performed by the first inference device 13 can be expanded to an appropriate second inference device 22 to perform the arithmetic processing. In other words, the first inference device 13 of the first intelligent device 10 processes the first partial information using the first inference model, and the second partial information is distributed to the second inference device 22 of the second intelligent device 20, so that the remaining second partial information can be expanded to be processed using the second inference model. In other words, the arithmetic processing of the processing information to be processed by the control system 1 can be expanded to the optimal second intelligent device 20 to perform the arithmetic processing.

[0066] Therefore, according to the control system 1, it is possible to optimize the calculation process using the inference model. As a result, it is possible to prevent the time required for calculating the processing information from becoming longer.

[0067] In the control system 1 according to the present embodiment, the distributed processing setting unit 30 includes a device information collecting unit 31 that collects one or more second intelligent devices 20 that can communicate with each other, a self-device processing estimating unit 32 that estimates the processing capacity of the first intelligent device 10 and outputs a first result including the estimated processing capacity, a candidate device processing estimating unit 33 that estimates the processing capacity of a candidate second intelligent device 20 among the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity, and a distributed configuration arranging unit 34. The candidate device processing estimating unit 33 determines whether it is appropriate to have the second inference device 22 perform the arithmetic processing of the second partial information based on the first result and the second result. Then, when it is determined that it is appropriate to have the second inference device 22 perform the arithmetic processing of the second partial information, the distributed configuration arranging unit 34 causes the second inference device 22 to perform the arithmetic processing of the second partial information using the second inference model.

[0068] According to this, the second inference device 22 that has been determined to be appropriate for performing the calculation process is made to perform the calculation process, so that the calculation process of the second partial information can be distributed to the second inference device 22. Therefore, it is possible to further optimize the calculation process by using the inference model.

[0069] Furthermore, in the control system 1 according to this embodiment, the candidate device processing estimation unit 33 estimates the processing capabilities of the second intelligent devices 20, generates a plurality of second results including the estimated processing capabilities, and determines the appropriateness of having the second inference device 22 perform arithmetic processing on the second partial information based on the first result and the plurality of second results. Then, when it is determined that it is appropriate to have the second inference device 22 perform arithmetic processing on the second partial information, the distributed configuration arrangement unit 34 causes the second inference device 22 to perform arithmetic processing on the second partial information using the second inference model.

[0070] According to this, the arithmetic processing is performed by a plurality of second inference devices 22 that have been determined to be appropriate for performing the arithmetic processing, so that the arithmetic processing of a plurality of pieces of second partial information can be distributed to the second inference devices 22. Therefore, it is possible to further optimize the arithmetic processing by using a plurality of inference models.

[0071] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capacity of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model, based on the spare processing capacity of the second inference device 22.

[0072] According to this, the second partial information can be processed by the second inference device 22 of the second intelligent device 20 having spare capacity, using the second inference model. In other words, a second inference device 22 suitable for processing the second partial information can be selected, and the selected second inference device 22 can be processed. This makes it possible to prevent a second intelligent device 20 having no spare capacity for processing from being selected and processing the second partial information using the second inference model. As a result, it is possible to prevent a decrease in the processing speed in the control system 1.

[0073] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capability of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capability. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model, based on the similarity between the function of the first inference device 13 and the function of the second inference device 22.

[0074] According to this, the second partial information can be processed by the second inference device 22 of the second intelligent device 20, which has a function similar to that of the first inference device 13, using the second inference model. In other words, the second inference device 22, which is suitable for the function of the first inference device 13, can be processed by the second inference device 22. This makes it possible to prevent the second inference device 22, which has a function dissimilar to that of the first inference device 13, from being selected and the second partial information being processed by using the second inference model. As a result, it is possible to prevent a decrease in the processing speed in the control system 1.

[0075] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capacity of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model based on the type of accelerator constituting the second inference device 22 using the second inference model and the type of operator constituting the second inference model.

[0076] This makes it possible to have the second inference device 22 using the second inference model having an accelerator type suitable for processing the second partial information using the second inference model perform the arithmetic processing. In other words, it becomes possible to select a second inference device 22 suitable for processing the second partial information, and have the selected second inference device 22 perform the arithmetic processing. This makes it possible to prevent the second inference device 22 having an accelerator type that is not suitable for processing the second partial information from performing the arithmetic processing. As a result, it is possible to prevent a decrease in the arithmetic processing speed in the control system 1.

[0077] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capacity of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model based on the type of accelerator constituting the second inference device 22 that uses the second inference model and the size of the first inference model.

[0078] This makes it possible to have the second inference device 22 using a second inference model suitable for the size of the first inference model process the second partial information. In other words, it becomes possible to select a second inference device 22 suitable for processing the second partial information, and have the selected second inference device 22 process the second partial information. This makes it possible to prevent the second inference device 22 using a second inference model that is smaller than the size of the first inference model and is not suitable for processing the second partial information from processing the second partial information. As a result, it is possible to prevent a decrease in the processing speed in the control system 1.

[0079] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capacity of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model based on whether or not the communication speed between the first intelligent device 10 and the second intelligent device 20 is faster than a reference value.

[0080] According to this, it becomes possible to have the second inference device 22 of the second intelligent device 20, in which the communication speed between the first intelligent device 10 and the second intelligent device 20 is faster than a reference value, perform the arithmetic processing. In other words, it becomes possible to select a second inference device 22 suitable for performing the arithmetic processing of the second partial information, and to have the selected second inference device 22 perform the arithmetic processing. Therefore, it is possible to prevent the second intelligent device 20, in which the communication speed between the first intelligent device 10 and the second intelligent device 20 is slower than the reference value, from being selected. As a result, it is possible to prevent a decrease in the arithmetic processing speed in the control system 1.

[0081] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capability of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capability. Then, based on the inference speed of the one or more second intelligent devices 20, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model.

[0082] This makes it possible to cause the second inference device 22 of the second intelligent device 20, which has a fast inference speed, to perform arithmetic processing of the second partial information using the second inference model. In other words, it becomes possible to select a second inference device 22 suitable for performing arithmetic processing of the second partial information, and cause the selected second inference device 22 to perform arithmetic processing. This makes it possible to prevent the second inference device 22 of the second intelligent device 20, which has an inference speed slower than the reference speed, from performing arithmetic processing. As a result, it is possible to prevent a decrease in the arithmetic processing speed in the control system 1.

[0083] Furthermore, in the control system 1 according to this embodiment, the distributed processing setting unit 30 has a candidate device processing estimation unit 33 that estimates the processing capacity of a candidate second intelligent device 20 from the collected one or more second intelligent devices 20 and generates a second result including the estimated processing capacity. Then, the candidate device processing estimation unit 33 determines the appropriateness of having the second inference device 22 perform arithmetic processing of the second partial information using the second inference model, based on the energy consumption of the one or more second intelligent devices 20.

[0084] This makes it possible to cause the second inference device 22 of the second intelligent device 20 with low energy consumption to perform arithmetic processing of the second partial information using the second inference model. In other words, it becomes possible to select a second inference device 22 suitable for performing arithmetic processing of the second partial information, and cause the selected second inference device 22 to perform arithmetic processing. This makes it possible to prevent the second inference device 22 of the second intelligent device 20 with higher energy consumption than a reference value from performing arithmetic processing. As a result, it is possible to prevent a decrease in energy efficiency in the control system 1.

[0085] In addition, in the control system 1 of this embodiment, when it is determined that it is appropriate to have the second inference device 22 perform calculation processing of the second partial information using the second inference model, the distributed configuration arrangement unit 34 distributes the second partial information to one or more second inference devices 22.

[0086] According to this, the second partial information can be distributed to one or more second inference devices 22 using the second inference model, so that a decrease in the calculation processing speed in the control system 1 can be suppressed.

[0087] In the control system 1 according to the present embodiment, the first intelligent device 10 further includes at least one of a sensor (first sensor 11) and an actuator (first actuator 12). The first inference device 13 controls the actuator (first actuator 12) according to the result of arithmetic processing based on the processing information output by the sensor (first sensor 11).

[0088] According to this, by suppressing an increase in the time required for calculating and processing the processing information, delays in the control of the first actuator 12 can be suppressed, and the state of the space can be changed in an optimal manner.

[0089] (Other variations) Although the control system according to the present disclosure has been described based on the above-mentioned embodiments, the present disclosure is not limited to these embodiments. As long as the modifications do not deviate from the spirit of the present disclosure, modifications that are conceivable by those skilled in the art may also be included in the scope of the present disclosure.

[0090] For example, with reference to Figs. 4 to 6, the first intelligent device 10, 10a, and the distributed processing setting unit 30 according to the above embodiment may be configured as shown in Figs. 4 to 6. Fig. 4 is a block diagram showing a first intelligent device 10a according to another modified example. Fig. 5 is a block diagram showing a device group 1c according to another modified example. Fig. 6 is a block diagram showing a cloud server 1d, an edge computer 1e, a first intelligent device 10, and a second intelligent device 20 according to another modified example.

[0091] The distributed processing setting unit 30 may be implemented in one or more first intelligent devices 10a as shown in Fig. 4. That is, the first intelligent device 10a may have at least one of the first sensor 11 and the first actuator 12, the first inference device 13, the first communication unit 14, and the distributed processing setting unit 30. Note that Fig. 4 illustrates an example in which the first intelligent device 10a has both the first sensor 11 and the first actuator 12.

[0092] 5, the distributed processing setting unit 30 may function as distributed processing for a device group 1c including a first intelligent device 10 and one or more second intelligent devices 20. The device group 1c shown in FIG. 5 may have a plurality of first intelligent devices 10, a plurality of second intelligent devices 20, and the distributed processing setting unit 30.

[0093] 6, the distributed processing setting unit 30 may be disposed in a cloud server 1d capable of communicating with the control system 1. A plurality of first intelligent devices 10 and a plurality of second intelligent devices 20 may be connected to the cloud server 1d via an edge computer 1e so as to be capable of communicating with the cloud server 1d.

[0094] Furthermore, in the control system 1 according to the above embodiment, when it is determined that it is appropriate to have multiple second inference devices 22 perform the arithmetic processing, the second partial information may be divided according to each of the multiple second inference devices 22 and distributed to each of the multiple second inference devices 22. Therefore, the number of second inference devices 22 that are determined to be appropriate is not limited to one.

[0095] Furthermore, the first inference device and the second inference device according to the above-described embodiments are typically realized as an LSI, which is an integrated circuit. These may be implemented individually as single chips, or may be implemented in a single chip so as to include some or all of them.

[0096] The integrated circuit is not limited to LSI, but may be realized by a dedicated circuit or a general-purpose processor. A field programmable gate array (FPGA) that can be programmed after LSI manufacturing, or a reconfigurable processor that can reconfigure the connections and settings of circuit cells inside the LSI may also be used.

[0097] In each of the above embodiments, each component of the first inference device, the second inference device, etc. may be configured with dedicated hardware, or may be realized by executing a software program suitable for each component. Each component may be realized by a program execution unit such as a CPU or a processor reading and executing a software program recorded on a storage medium such as a hard disk or a semiconductor memory.

[0098] Furthermore, all the numbers used above are merely examples for the purpose of specifically explaining the present disclosure, and the embodiments of the present disclosure are not limited to the numbers exemplified.

[0099] In addition, the division of functional blocks in the block diagram is an example, and multiple functional blocks may be realized as one functional block, one functional block may be divided into multiple blocks, some functions may be transferred to other functional blocks, and the functions of multiple functional blocks having similar functions may be processed in parallel or in a time-sharing manner by a single piece of hardware or software.

[0100] In addition, the order in which each step is performed in the flowchart is merely for illustrative purposes and may be other than the above. In addition, some of the steps may be performed simultaneously (in parallel) with other steps.

[0101] In addition, the present disclosure also includes forms obtained by applying various modifications to the above-described embodiments that a person skilled in the art may conceive, and forms realized by arbitrarily combining the components and functions of each embodiment within the scope that does not deviate from the spirit of the present disclosure.

[0102] (Additional Note) The following describes the features of the control system described based on the above embodiments.

[0103] <Technology 1> a first intelligent device having a first inference device that uses a first inference model to perform computational processing of processing information and a communication unit; A distributed processing setting unit, the first inference device performs computational processing on first partial information, which is a part of the processing information, using the first inference model; the communication unit is capable of communicating with one or more second intelligent devices that are devices other than the first intelligent device; The distributed processing setting unit extracting candidates for the second intelligent device having a second inference device that performs arithmetic processing using a second inference model from among the one or more second intelligent devices; Using the second inference model, the appropriateness of having the second inference device in the extracted candidate second intelligent device perform arithmetic processing on second partial information, which is the remaining processing information of the processing information, is determined. Control system.

[0104] <Technology 2> The distributed processing setting unit A device information collection unit that collects one or more of the second intelligent devices that can communicate with the device information collection unit; a self-device processing estimation unit that estimates a processing capability of the first intelligent device and outputs a first result including the estimated processing capability; a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; A distributed configuration arrangement unit, the candidate device processing estimation unit determines, based on the first result and the second result, whether it is appropriate to have the second inference device perform arithmetic processing on the second partial information; When it is determined that it is appropriate to cause the second inference device to perform arithmetic processing on the second partial information, the distributed configuration arrangement unit causes the second inference device to perform arithmetic processing on the second partial information using the second inference model. The control system described in technique 1.

[0105] <Technology 3> The candidate device processing estimation unit Estimating processing capabilities of a plurality of the second intelligent devices and generating a plurality of second results including the estimated processing capabilities; determining whether it is appropriate to cause the second inference device to perform a calculation process on the second partial information based on the first result and the second results; When it is determined that it is appropriate to cause the second inference device to perform arithmetic processing on the second partial information, the distributed configuration arrangement unit causes the second inference device to perform arithmetic processing on the second partial information using the second inference model. The control system described in technique 2.

[0106] <Technology 4> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a spare processing capacity of the second inference device. The control system according to any one of the first to third aspects.

[0107] <Technology 5> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a similarity between a function of the first inference device and a function of the second inference device. The control system according to any one of the first to fourth aspects.

[0108] <Technology 6> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a type of accelerator constituting the second inference device that uses the second inference model and a type of operator constituting the second inference model. The control system according to any one of the first to fifth aspects.

[0109] <Technology 7> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a type of accelerator constituting the second inference device that uses the second inference model and a size of the first inference model. The control system according to any one of the first to sixth aspects.

[0110] <Technology 8> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model based on whether or not a communication speed between the first intelligent device and the second intelligent device is faster than a reference value. The control system according to any one of the first to seventh aspects.

[0111] <Technology 9> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on an inference speed of one or more of the second intelligent devices. The control system according to any one of the first to eighth aspects.

[0112] <Technology 10> the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on the energy consumption of one or more of the second intelligent devices. The control system according to any one of the first to ninth aspects of the present invention.

[0113] <Technology 11> the distributed processing setting unit includes a distributed configuration arrangement unit that distributes the second partial information to one or more of the second inference devices when it is determined that it is appropriate to have the second inference device perform arithmetic processing of the second partial information using the second inference model. The control system according to any one of the first to tenth aspects.

[0114] <Technology 12> The distributed processing setting unit is implemented in one or more of the first intelligent devices. The control system according to any one of the first to eleventh aspects of the present invention.

[0115] <Technology 13> The distributed processing setting unit functions as distributed processing for a device group including the first intelligent device and one or more of the second intelligent devices. The control system according to any one of the first to twelfth aspects.

[0116] <Technology 14> The distributed processing setting unit is disposed in a cloud server capable of communicating with the control system. The control system according to any one of the first to thirteenth aspects.

[0117] <Technology 15> the first intelligent device further comprises at least one of a sensor and an actuator; The first inference device controls the actuator in response to a result of arithmetic processing based on the processing information output by the sensor. The control system according to any one of the first to fourth aspects of the present invention. [Explanation of symbols]

[0118] 1. Control System 1c Device Group 1d Cloud Server 10, 10a First intelligent device 11 First sensor (sensor) 12 First actuator (actuator) 13 First inference device 14 First Communication Department (Communication Department) 20. Second Intelligent Device 22 Second Reasoning Device 30 Distributed processing setting section 31 Device Information Collection Department 32 Own device processing estimation unit 33 Candidate device processing estimation unit 34 Distributed configuration arrangement part

Claims

1. a first intelligent device having a first inference device that uses a first inference model to perform computational processing of processing information and a communication unit; A distributed processing setting unit, the first inference device performs computational processing on first partial information, which is a part of the processing information, using the first inference model; the communication unit is capable of communicating with one or more second intelligent devices that are devices other than the first intelligent device, The distributed processing setting unit extracting candidates for the second intelligent device having a second inference device that performs arithmetic processing using a second inference model from the one or more second intelligent devices; Using the second inference model, it is determined whether it is appropriate to have the second inference device in the extracted candidate second intelligent device perform arithmetic processing on second partial information, which is the remaining processing information of the processing information. Control system.

2. The distributed processing setting unit a device information collection unit that collects one or more of the second intelligent devices that can communicate with the device information collection unit; a self-device processing estimation unit that estimates a processing capability of the first intelligent device and outputs a first result including the estimated processing capability; a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; A distributed configuration arrangement unit, the candidate device processing estimation unit determines, based on the first result and the second result, whether it is appropriate to have the second inference device perform arithmetic processing on the second partial information; When it is determined that it is appropriate to cause the second inference device to perform arithmetic processing on the second partial information, the distributed configuration arrangement unit causes the second inference device to perform arithmetic processing on the second partial information using the second inference model. The control system of claim 1 .

3. The candidate device processing estimation unit Estimating processing capabilities of a plurality of the second intelligent devices and generating a plurality of second results including the estimated processing capabilities; determining whether it is appropriate to cause the second inference device to perform a calculation process on the second partial information based on the first result and the second results; When it is determined that it is appropriate to cause the second inference device to perform arithmetic processing on the second partial information, the distributed configuration arrangement unit causes the second inference device to perform arithmetic processing on the second partial information using the second inference model. The control system of claim 2.

4. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a spare processing capacity of the second inference device. A control system according to any one of claims 1 to 3.

5. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a similarity between a function of the first inference device and a function of the second inference device. A control system according to any one of claims 1 to 3.

6. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a type of accelerator constituting the second inference device that uses the second inference model and a type of operator constituting the second inference model. A control system according to any one of claims 1 to 3.

7. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on a type of accelerator constituting the second inference device that uses the second inference model and a size of the first inference model. A control system according to any one of claims 1 to 3.

8. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model based on whether or not a communication speed between the first intelligent device and the second intelligent device is faster than a reference value. A control system according to any one of claims 1 to 3.

9. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on an inference speed of one or more of the second intelligent devices. A control system according to any one of claims 1 to 3.

10. the distributed processing setting unit includes a candidate device processing estimation unit that estimates a processing capability of a candidate second intelligent device from among the collected one or more second intelligent devices, and generates a second result including the estimated processing capability; The candidate device processing estimation unit determines the appropriateness of having the second inference device perform arithmetic processing of the second partial information using the second inference model, based on the energy consumption of one or more of the second intelligent devices. A control system according to any one of claims 1 to 3.

11. When it is determined that it is appropriate to cause the second inference device to perform arithmetic processing of the second partial information using the second inference model, the distributed configuration and arrangement unit distributes the second partial information to one or more of the second inference devices. A control system according to claim 2 or 3.

12. The distributed processing setting unit is implemented in one or more of the first intelligent devices. A control system according to any one of claims 1 to 3.

13. The distributed processing setting unit functions as a distributed processing unit for a device group including the first intelligent device and one or more of the second intelligent devices. A control system according to any one of claims 1 to 3.

14. The distributed processing setting unit is disposed in a cloud server capable of communicating with the control system. A control system according to any one of claims 1 to 3.

15. The first intelligent device further includes at least one of a sensor and an actuator. The first inference device controls the actuator in response to a result of arithmetic processing based on the processing information output by the sensor. A control system according to any one of claims 1 to 3.

Images