Information processing system, information processing method, and program

Abstract

To improve convenience of management of a facility that is managed based on data from a plurality of sensors or cameras.SOLUTION: In an information processing system includes a server 1 that executes predetermined processing using input data, and one or more input cells 2 that provide at least part of the input data to the server 1, the input cells 2 acquire information on the positions of the input cells. The server 1 manages input cell position information and a model in a three-dimensional map of a facility in which the input cells 2 are arranged. The input cells 2 execute re-learning based on information on the three-dimensional map and information on the positions of the input cells 2 managed by the server 1.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to an information processing system, an information processing method, and a program.

Background Art

[0002] Conventionally, in a premises (for example, inside a factory), there is a technique of controlling various devices based on parameters measured by various sensors to change the parameters and manufacture better products (for example, see Patent Document 1). In recent years, the technology of artificial intelligence (hereinafter referred to as "AI") has been developing. For example, AI capable of natural language processing, which is used in chatbots and the like, has also been developing.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the detection of failures and prediction of failures from the operating status of various devices used in a premises, conventionally, various sensors have been centrally detected, judged, and alarm outputs have been performed by software in a central management device. However, when an accident or the like occurs and data (abnormal values) is transmitted from many sensors at once, in the central monitoring device, since sequential processing is performed, it takes a considerable amount of time until an alarm output to a user (for example, a monitor) or an operation of emergency equipment (for example, fire extinguishing by a sprinkler) is performed, and the response may be delayed, leading to a major disaster. Therefore, a system that can start a response in a shorter time or can be used without constructing a large centralized system has been desired.

[0005] The present invention can improve the convenience of managing equipment that is managed based on data from multiple sensors or cameras. [Means for solving the problem]

[0006] To achieve the above objective, an information processing system according to one aspect of the present invention is: An information processing system including a central device that performs predetermined processing using input data, and one or more Type 1 peripheral devices that provide at least a portion of the input data to the central device, The aforementioned central device is A processing execution means that acquires one or more perceptual representation data as input data, performs a predetermined process using the input data, and outputs one or more perceptual representation data indicating the result of the process as output data. Equipped with, Each of the one or more Type 1 peripheral devices is: A model management means that takes predetermined data as input, converts it into the aforementioned perceptual representation data, and outputs a model, which is then stored and managed in a predetermined storage medium. A conversion means that acquires data output from a sensor that measures physical quantities in the real world, or from a camera that images a target area, inputs it into the model, and outputs the perceptual representation data output from the model as at least a part of the input data of the central device, It is equipped with.

[0007] An information processing method and program according to one aspect of the present invention are an information processing method and program corresponding to the information processing system according to one aspect of the present invention described above. [Effects of the Invention]

[0008] According to the present invention, the convenience of managing equipment that is managed based on data from multiple sensors or cameras can be improved. [Brief explanation of the drawing]

[0009] [Figure 1]This is a schematic diagram illustrating an example of a service to which an "autonomous decentralized AI blockchain cell" is applied. [Figure 2] This figure shows an example configuration of an information processing system according to one embodiment of the present invention, which is applied when providing the service shown in Figure 1. [Figure 3] Figure 2 is a block diagram showing an example of the server hardware configuration in the information processing system. [Figure 4] Figure 3 is a functional block diagram showing an example of the functional configuration of an information processing system including a server with the hardware configuration shown. [Figure 5] This figure shows an example of using the service shown in Figure 1 for managing equipment located within a premises. [Figure 6] Figure 5 shows an example of the information processing flow for high-speed handling within the campus. [Figure 7] This figure shows an example of the process for autonomously establishing the processing content in the input cell of the functional configuration shown in Figure 4. [Figure 8] This is a schematic diagram illustrating an example of a service to which the "autonomous decentralized AI blockchain audiovisual cell" will be applied. [Figure 9] Figure 8 is a functional block diagram showing an example of the functional configuration of an information processing system that provides the service. [Figure 10] Figure 7 shows an example of using the service shown to manage equipment located within the premises. [Figure 11] Figure 10 shows an example of the information processing flow for high-speed handling within the campus. [Modes for carrying out the invention]

[0010] Embodiments of the present invention will be described below with reference to the drawings.

[0011] In the following, the term "image" includes both "moving images" and "still images." Furthermore, "moving images" shall include images displayed by each of the following first to third processes. The first process refers to a process of continuously switching and displaying a series of still images consisting of multiple images over time for each movement of an object (for example, an image of a person or an object to be imaged) in a planar image (2D image). Specifically, for example, 2D animation, so-called para para manga-like processing corresponds to the first process. The second process refers to a process of setting motions corresponding to the respective motions of objects (for example, images of persons or objects to be imaged) in a stereoscopic image (image of a 3D model), and changing and displaying the motions over time. Specifically, for example, 3D animation corresponds to the second process. The third process refers to a process of preparing videos (i.e., moving images) corresponding to the respective motions of objects (for example, images of natural persons) and playing the videos over time. Here, the "video (i.e., moving image)" is composed of images such as a plurality of frames and fields (hereinafter referred to as "unit images").

[0012] Hereinafter, the "autonomous decentralized AI blockchain cell" and the "autonomous decentralized AI blockchain audiovisual cell", which are the premises of the present invention, will be described with reference to FIGS. 1 to 11.

[0013] Although it will be described in detail later, the difference between the "autonomous decentralized AI blockchain cell" and the "autonomous decentralized AI blockchain audiovisual cell" will be briefly described here. This service uses blockchain technology in which, in a facility (equipment, improvement, etc.), in response to an input from an input cell (for example, input cells 2-1 and 2-2 in FIG. 1), an output by predetermined control is made from an output cell (for example, output cell 3-1 in FIG. 1).

[0014] Specifically, the "autonomous decentralized AI blockchain cell" is provided with a sensor for measuring a physical quantity in the real world (for example, temperature) in an input cell (for example, input cell 2-1 in FIG. 1), and a plurality of such cells are provided in the facility and operate autonomously. In contrast, an "autonomous decentralized AI blockchain audiovisual cell" is an input cell (for example, input cell 2-1 in Figure 8) equipped with a camera and microphone that capture images of a predetermined area, and multiple such cells are installed within the facility and operate autonomously.

[0015] The "autonomous decentralized AI blockchain cell" will be explained below using Figures 1 to 7, and the "autonomous decentralized AI blockchain audiovisual cell" will be explained using Figures 8 to 11. Figure 1 is a schematic diagram illustrating an example of a service to which an "autonomous decentralized AI blockchain cell" is applied (hereinafter referred to as "this service"). This service, to which the "autonomous decentralized AI blockchain cell" is applied, utilizes the information processing system shown in Figure 1 to provide a system that outputs the operation of a control device based on the input of data measured by multiple sensors.

[0016] The information processing system for this service, as shown in Figure 1, consists of Server 1, input cells 2-1 and 2-2, output cell 3-1, verification terminal 4, and blockchain full nodes 5-1 to 5-4 (hereinafter referred to as "BC full nodes" as shown in Figure 1). Furthermore, as will be explained in more detail later, input cells 2-1 and 2-2, and output cell 3-1 are equipped with the functionality of blockchain ultralight nodes (hereinafter referred to as "BC ultralight nodes" as shown in Figure 1).

[0017] Server 1 is an information processing device equipped with a central AI 61. As will be described in more detail later, Server 1 acquires sensor measurement results from input cells 2-1 and 2-2 converted into perceptually represented data as input data, and outputs perceptually represented control instructions from output cell 3-1 based on the input data as output data.

[0018] Input cells 2-1 and 2-2 are equipped with a sensor and an autonomous distributed chip CP, respectively. The measurement results from the sensor are converted into perceptually represented data by the autonomous distributed chip CP and managed using the blockchain network BCN.

[0019] Output cell 3-1 is equipped with an autonomous distributed chip CP and an actuator (an example of a control device). The autonomous distributed chip CP receives control instructions output by server 1 from the blockchain network BCN and controls the control device (actuator) based on the control instructions.

[0020] Verification terminal 4 is an information processing device for verifying information managed using the blockchain network BCN. Equipment managers (such as those at factories) can use verification terminal 4 to verify the history of inputs and outputs that have not been tampered with.

[0021] Here, we will explain "perceived data." The raw data from sensor measurements is numerical data, representing physical quantities. In contrast, humans can interpret such sensor measurement results and express the results using language or other means. Specifically, for example, humans can perceive (recognize, interpret) raw data such as a temperature of 35 degrees Celsius and humidity of 80% as "hot" and express it in words. This kind of expression, which is not the raw data of the sensor's measurement results but rather a perceived (recognized, interpreted) expression, is called a "perceptual expression." Furthermore, the form of "language" is just one example of a form of perceptual representation. That is, for example, the data of the string "hot" is data that is perceptually represented in the form of the Japanese language. Also, for example, if an identifier is pre-assigned in an information processing device, the data of the identifier corresponding to "hot" is also an example of perceptually represented data that can be used within the information processing device. Moreover, for example, video data of the action of fanning one's face with one's hand can also be said to be data that is perceptually represented in the form of a gesture. For the sake of simplicity, the following explanation will assume that "perceptual expression" is "language."

[0022] Furthermore, as mentioned above, this service utilizes blockchain technology. Generally, the term "blockchain" can refer to distributed ledger technology or a distributed network. In other words, the term blockchain is a polysemous term that includes not only the data itself, which is a series of data linked together like a chain called a block, but also the technology and network related to it. Therefore, below, the decentralized network that manages the blockchain will be referred to as the "blockchain network," distinguishing it from the "blockchain," which is a series of data where blocks are linked together like a chain. In other words, a "blockchain" is a series of data where "blocks" are linked together like a chain, each containing various information (such as the data itself, metadata, and data related to verifying integrity, such as hash values) about one or more data managed using this service (for example, data based on sensor measurements or data for controlling devices).

[0023] First, we will explain the functions of the various nodes in the BCN blockchain network used in this service.

[0024] The blockchain network BCN consists of multiple nodes, with at least one node residing in the cloud. In the example shown in Figure 1, each of the four BC full nodes 5-1 through 5-4 functions as a full node on the cloud. Here, a "full node" is an information processing device (node) that provides all the functions of a node in a blockchain, such as the computational processing functions related to block generation and the storage functions for the blockchain data itself. BC full nodes 5-1 through 5-4 form a blockchain network (BCN) that communicates with each other.

[0025] In the example shown in Figure 1, the BC ultralight nodes provided in the two input cells 2-1 and 2-2, and the output cell 3-1, respectively, are all functional. Here, an "ultralight node" is an information processing device (node) that does not provide computational processing functions related to block generation or storage functions for the blockchain data itself, but rather provides only a very limited set of functions, such as data exchange functions with the blockchain network BCN. Because ultralight nodes require fewer computing resources to implement, they are implemented as part or all of the chip or program that provides the aforementioned functions. However, ultralight nodes may also be implemented as separate information processing devices.

[0026] Although not shown in the diagram, a node may include a "light node." Here, a "light node" is a node that does not function as a full node but is responsible for some of the computational processing functions related to block generation and the storage functions of the blockchain data itself.

[0027] The BC ultralight nodes located in input cells 2-1 and 2-2, and output cell 3-1, communicate with BC full nodes 5-1 through 5-4 by connecting to the cloud via a dedicated line. In other words, in the example shown in Figure 1, the blockchain network BCN is formed by a total of seven nodes: four BC full nodes 5-1 to 5-4 and three BC ultralight nodes, each functioning as a separate node.

[0028] Here, a dedicated line refers to a communication line exclusively for a specific user. For example, communication over a dedicated line is isolated from networks (such as the internet) that include untrusted information processing devices. In other words, communication between information processing devices connected via a dedicated line is less likely to be eavesdropped on or intercepted by malicious third parties. That is, because this service is used via a dedicated line, it is provided in a state where the possibility of eavesdropping or intercepting by third parties is low. It should be noted that the dedicated line does not necessarily have to be physically isolated from the internet, etc. For example, a virtual dedicated line using VPN (Virtual Private Network) technology can also be adopted as the dedicated line described above.

[0029] The above illustrates an example of the configuration of the information processing system in this service to which the "autonomous distributed AI blockchain cell" is applied, using Figure 1. The following details of the information processing flow in this service will be explained using Figure 1, following steps ST11 to ST19.

[0030] In step ST11, numerical data indicating that the temperature is rising and humidity is rising are obtained as measurement results from the sensor in input cell 2-1. Specifically, for example, numerical data (digital signal) indicating that the temperature is rising from 20 degrees to 40 degrees over time, and numerical data (digital signal) indicating that the humidity is rising from 50% to 80% over time are obtained.

[0031] In step ST12, the language conversion AI of the autonomous distributed chip CP in input cell 2-1 converts digital signals into language. Specifically, for example, numerical data (digital signals) indicating that the temperature is rising from 20 degrees to 40 degrees over time, or numerical data (digital signals) indicating that the humidity is rising from 50% to 80% over time, are converted into the language data "hot".

[0032] In step ST13, the BC ultralight node of the autonomous distributed chip CP in input cell 2-1 manages the language data "hot" using the blockchain network BCN. Specifically, for example, it manages the language data "hot" by sending it to the BC full node 5-1 as part of a transaction in blockchain technology. At this time, the BC ultralight node of the autonomous distributed chip CP in input cell 2-1 manages the encrypted language data using the blockchain network BCN.

[0033] Although not shown in the diagram, the processes described in steps ST11 to ST13 are also executed in input cell 2-2. Specifically, for example, data indicating the presence of multiple heat sources is obtained as a measurement result from the sensor in input cell 2-2. Specifically, for example, infrared thermography data is obtained as data indicating the presence of multiple heat sources. For example, the language conversion AI in the autonomous distributed chip CP of input cell 2-2 converts data indicating the presence of many heat sources into language data indicating "a large number of people." For example, the BC ultralight node in the autonomous distributed chip CP of input cell 2-2 manages encrypted language data, such as "many people," using the blockchain network BCN.

[0034] In step ST14, Server 1 decrypts and obtains encrypted language data managed using the blockchain network BCN. In step ST15, Server 1 uses the blockchain network BCN to manage the results of its determination of the content of control instructions based on the acquired language data. Specifically, for example, Server 1 uses Central AI 61, which has been trained as an AI capable of natural language processing, to generate the control instruction "Lower the temperature significantly overall" as output data, using the language data "hot" and "crowded" as input data. Server 1 then manages the generated control instruction "Lower the temperature significantly overall" using the blockchain network BCN. At this time, Server 1 manages the encrypted control instruction using the blockchain network BCN.

[0035] In step ST16, the BC ultralight node of the autonomous distributed chip CP in output cell 3-1 obtains encrypted control instructions managed using the blockchain network BCN.

[0036] In step ST17, the firmware of the autonomous distributed chip CP in output cell 3-1 decrypts the encrypted control instructions. Then, the language conversion AI of the autonomous distributed chip CP in output cell 3-1 converts them into digital signals as specific control content for the control device. Specifically, for example, the language conversion AI in the autonomous distributed chip CP of output cell 3-1 converts a control instruction to "lower the temperature more strongly overall" into a digital signal that causes the control device to "increase the output" of the air conditioner and set the "airflow direction to all directions". In other words, for example, the language conversion AI in the autonomous distributed chip CP of output cell 3-1 outputs a control signal (digital signal) of digital data that operates the actuator that controls the air conditioner. Although not shown in the diagram, the BC ultralight node on the autonomous distributed chip CP in output cell 3-1 can also manage the operation of the control device that "increases the output" of the air conditioner and "directs the airflow in all directions" using the blockchain network BCN.

[0037] In step ST18, the actuator (an example of a control device) is driven according to the control signals (digital signals) of the digital data that control the operation of the actuator that operates the air conditioner cooling, namely "increase output" and "general output." As a result, the actuator (an example of a control device) operates in accordance with the control instruction from the central AI61 to "lower the temperature more strongly overall." As a result, based on the measurement results from sensors such as "temperature rise," "humidity rise," and "multiple heat sources," the air conditioner's cooling output is increased and the output is distributed in all directions (not shown in the diagram), improving the environment.

[0038] In step ST19, the verification terminal 4 can present information managed using the blockchain network BCN to the facility (factory, etc.) manager. Specifically, the facility (factory, etc.) manager can verify the sensor measurement results such as "temperature rise," "humidity rise," and "multiple heat sources" acquired by the sensors, input data (linguistic data) such as "hot" and "many people," output data (linguistic data) of control instructions such as "lower the temperature more strongly overall," and the operation of the control device that "increases the output" of the air conditioner and sets the "airflow direction to all directions," as presented to the verification terminal 4. Since this information is managed using the blockchain network BCN, it is presented to the facility (factory, etc.) manager as an unaltered record of input and output history.

[0039] Due to the configuration and operation described above, this service has the following characteristics: Firstly, an existing AI that has been trained in language can be adopted as the central AI 61. In other words, this service allows the use of existing AIs that have been trained in language, acting as the central AI61, thus reducing overall development and operational costs. Specifically, when building such a system in new equipment, the sensors used and the underlying environment differ for each piece of equipment. Therefore, the values ​​that the sensors should measure also differ, and it is necessary to train the AI ​​according to the equipment. However, AI capable of natural language processing, which generates input data (linguistic data) such as "it's hot" and "there are a lot of people," and output data (linguistic data) for control instructions such as "lower the temperature overall more significantly," is being used for many purposes, and therefore various developments are underway. As a result, the accuracy of such AI capable of natural language processing is improving, and the cost is decreasing. Therefore, this service can reduce overall development and operational costs.

[0040] Secondly, because AI processing is performed on the edge side, the processing burden on the central AI61 is reduced. In other words, in this service, input cells 2-1 and 2-2 (edge ​​side) perform processing to convert the sensor measurement results into language data (an example of perceptual expression), and output cell 3-1 (edge ​​side) performs processing to convert the language data (output data of control instructions) into digital signals as specific control content for the control device. In other words, the central AI61 does not need to perform these processes, thus reducing its workload. This means that distributed processing of AI is being achieved. Furthermore, the effect of this distributed processing becomes more pronounced as the number of input and output cells on the edge side increases.

[0041] Thirdly, this service uses the blockchain network BCN to guarantee data integrity, thus realizing a secure AI system. In other words, for example, in a system that does not use the blockchain network BCN, it is possible that an incorrect control instruction may be output by impersonating input cell 2-1 and inputting linguistic data such as "temperature decrease" into the central AI 61. This service uses the blockchain network BCN to guarantee data, making it impossible to impersonate input cell 2-1, thus realizing a secure AI system.

[0042] The overview of this service was explained above using Figure 1. The following explanation, using Figures 2 through 4, describes the information processing system applied when providing the service shown in Figure 1 of this service. Figure 2 shows an example configuration of an information processing system according to one embodiment of the present invention, which is applied when providing the service shown in Figure 1.

[0043] In other words, the example of the information processing system configuration shown in Figure 2 is a more general system configuration of the information processing device for this service shown in Figure 1.

[0044] Server 1 is an information processing device managed by service provider S. Server 1 performs various processes to realize this service while communicating as appropriate with input cells 2-1 to 2-N (where N is an integer value of 1 or more), output cells 3-1 to 3-M (where M is an integer value of 1 or more independent of N), confirmation terminal 4, and BC full nodes 5-1 to 5-L (where L is an integer value of 1 or more independent of N and M).

[0045] Input cells 2-1 through 2-N are information processing devices equipped with one or more sensors and an autonomous distributed chip CP. When illustrating and explaining one of the N input cells 2-1 through 2-N, we will use the term "input cell 2-p" (where p is an integer value between 1 and N).

[0046] Output cells 3-1 to 3-M are information processing devices comprising one or more control devices and an autonomous distributed chip CP. When illustrating and explaining one of the M output cells 3-1 to 3-M, the term "output cell 3-q" (where q is an integer value between 1 and M) is used.

[0047] Verification terminal 4 is an information processing device operated by facility (factory, etc.) management personnel to verify information managed using the blockchain network BCN. Although one verification terminal 4 is shown in Figure 2, the number of verification terminals 4 is arbitrary.

[0048] BC full nodes 5-1 through 5-L are information processing devices (nodes) that provide all the functions of a blockchain node, such as computational processing functions related to block generation and storage functions for the blockchain data itself. In the example shown in Figure 2, BC full nodes 5-1 through 5-L reside on Cloud C.

[0049] Figure 3 is a block diagram showing an example of the server hardware configuration in the information processing system shown in Figure 2.

[0050] Server 1 comprises a CPU (Central Processing Unit) 11, ROM (Read Only Memory) 12, RAM (Random Access Memory) 13, a bus 14, an input / output interface 15, an input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0051] The CPU 11 executes various processes according to the program recorded in the ROM 12 or the program loaded from the storage unit 18 into the RAM 13. RAM13 also stores data and other information necessary for the CPU11 to perform various processes.

[0052] The CPU 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output interface 15 is also connected to this bus 14. An input / output interface 15 is connected to an input unit 16, an output unit 17, a storage unit 18, a communication unit 19, and a drive 20.

[0053] The input unit 16 is configured, for example, with a keyboard, and is used to input various types of information. The output unit 17 consists of a display such as an LCD and a speaker, and outputs various information as images and sounds. The memory unit 18 is composed of DRAM (Dynamic Random Access Memory) and stores various types of data. The communication unit 19 communicates with other devices (for example, input cells 2-1 to 2-N, output cells 3-1 to 3-M, confirmation terminal 4, and BC full nodes 5-1 to 5-L in Figure 2) via a network including the Internet.

[0054] A removable media 31, such as a magnetic disk, optical disk, magneto-optical disk, or semiconductor memory, is appropriately mounted in the drive 20. Programs read from the removable media 31 by the drive 20 are installed in the storage unit 18 as needed. Furthermore, the removable media 31 can store various types of data stored in the storage unit 18, just as the storage unit 18 does.

[0055] Although not shown in the diagram, input cells 2-1 to 2-N, output cells 3-1 to 3-M, verification terminal 4, and BC full nodes 5-1 to 5-L in Figure 2 can also have a configuration that is basically the same as the hardware configuration shown in Figure 3. Therefore, a description of the hardware configuration of input cells 2-1 to 2-N, output cells 3-1 to 3-M, verification terminal 4, and BC full nodes 5-1 to 5-L will be omitted. However, as shown in Figure 1, input cells 2-1 to 2-N are equipped with sensors as input units. Furthermore, input cells 2-1 to 2-N have some or all of the CPU, ROM, RAM, etc., as an autonomous distributed chip CP. Similarly, output cells 3-1 to 3-M are equipped with a control device (e.g., an actuator) as an output unit, as shown in Figure 1. Furthermore, output cells 3-1 to 3-M have some or all of the CPU, ROM, RAM, etc., as an autonomous distributed chip CP.

[0056] Through the collaboration of various hardware and software components of Server 1 as shown in Figure 3, various processes can be executed. As a result, the aforementioned service can be provided. The following describes the functional configuration of the information processing system, including Server 1 in Figure 3, as shown in Figure 2.

[0057] Figure 4 is a functional block diagram showing an example of the functional configuration of an information processing system including the server with the hardware configuration shown in Figure 3.

[0058] As shown in Figure 4, the CPU 11 of server 1 has a processing execution unit 51 that functions. The memory unit 18 stores the model of the central AI 61.

[0059] In input cell 2-p, the autonomous distributed chip CP includes a model management unit 71, a language conversion AI 72, a retraining execution unit 73, and a sensor 74. Input cell 2-p has the sensor 74 as its input unit. The language conversion AI model 75 is stored in the memory unit of input cell 2-p.

[0060] In output cell 3-q, the autonomous distributed chip CP functions as a model management unit 81, a reverse language conversion AI 82, a control unit 83, and a retraining execution unit 84. Output cell 3-1 has an actuator 85 as an output unit. In addition, the reverse language conversion AI model 86 is stored in the memory unit of output cell 3-q.

[0061] The processing execution unit 51 acquires input language data from one or more input cells 2-p as input data, executes a predetermined process using the input data, and outputs one or more output language data indicating the execution result of that process as output data. The processing execution unit 51 retrieves the input data stored using the blockchain network BCN.

[0062] The model management unit 71 takes a signal indicating the measurement result from the sensor 74 as input, converts it into language data, and stores and manages a language conversion AI model 75 in the memory unit, which then outputs the result. Specifically, for example, the model management unit 71 converts measurement results such as temperature, humidity, and thermography into linguistic data such as "temperature rising," "humidity rising," and "many people" and outputs them.

[0063] The language conversion AI 72 acquires a signal indicating a predetermined physical quantity output from a sensor 74 that detects a predetermined physical quantity in the real world, inputs it to the language conversion AI model 75, and outputs the language data output from the model as at least part of the input language data of the server 1.

[0064] The retraining execution unit 73 performs retraining to update the model.

[0065] The BC ultralight node in input cell 2-p stores at least a portion of the input post-language data output from language translation AI72 using the blockchain network BCN.

[0066] The model management unit 81 takes language data as input, converts it into predetermined physical quantities, and outputs an inverse language conversion AI model 86, which is then stored and managed in a predetermined storage medium.

[0067] The BC ultralight node in output cell 3-q retrieves the stored output language data using the blockchain network BCN and provides it to the reverse language conversion AI82. The reverse language conversion AI 82 acquires at least a portion of one or more linguistic representation data that constitute the output data output from server 1 and inputs it into the reverse language conversion AI model 86, which then outputs a signal indicating a physical quantity.

[0068] The control unit 83 controls the actuator 85 by inputting a signal indicating a physical quantity output from the model as an instruction signal to the actuator 85.

[0069] The retraining execution unit 84 performs retraining to update the model.

[0070] Examples of relearning performed by the relearning execution unit 73 and the relearning execution unit 84 will be described later. With this functional configuration, the information processing system for this service can perform each of the processes described using Figure 1, etc.

[0071] Furthermore, the information processing system of this embodiment can be utilized as described below. The following examples illustrate the application of this information processing system to monitoring multiple devices located within a premises, using Figures 5 and 6. Figure 5 shows an example of using the service shown in Figure 1 to manage equipment located within the premises. In the example shown in Figure 5, four devices A1 to A4 are located within the premises. That is, for example, devices A1 to A4 are each piece of equipment on a manufacturing line located within the factory premises.

[0072] Furthermore, an input cell 2-1 is located in device A1. This illustrates, for example, how a certain event related to device A1 on the manufacturing line is measured by the sensor in input cell 2-1. Specifically, the sensor measures arbitrary physical quantities, such as the temperature at a predetermined location on device A1 or the weight of materials placed into device A1.

[0073] Furthermore, device A2 is equipped with input cells 2-2 and 2-3. This shows how two physical quantities of device A2 are measured by the sensors of the two input cells 2-2 and 2-3. In other words, while device A1 measured the physical quantities of one device A1 with the sensor of one input cell 2-1, multiple input cells can be arranged for a single device.

[0074] Furthermore, input cells 2-4 are located inside device A3. This indicates that input cells 2-4 are built into device A3, and that the sensors in input cells 2-4 measure the physical quantities of device A3.

[0075] Furthermore, input cells 2-5 and 2-6 are located inside device A4. This indicates that the two input cells 2-5 and 2-6 are built into a single device A4, and that the sensors in the two input cells 2-5 and 2-6 measure the physical quantities of device A4.

[0076] Then, each input cell 2-1 to 2-6 converts the data measured by the sensor into language data and transmits it to the edge server EDS. In this way, data from multiple devices A1 to A4 within the premises is collected.

[0077] This enables fault detection and prediction based on the operating status of various devices used within the premises (e.g., manufacturing robots (lines), room temperature control, lighting control, etc.). Furthermore, it enables unmanned monitoring of the premises (e.g., factories, hospitals, apartment buildings, etc.) as shown in Figure 5. As mentioned above, in conventional monitoring systems, when an accident occurs and data (abnormal values) is transmitted simultaneously from many sensors, the central monitoring device processes the data sequentially, which can result in a considerable delay before it can issue an alarm to the user (e.g., a monitoring officer) or activate emergency equipment (e.g., fire suppression using sprinklers).

[0078] In contrast, this information processing system distributes processing because each input cell is converted into language data, enabling high-speed processing overall.

[0079] Furthermore, by processing without going through control instructions from the central AI61, it is possible to handle the situation even faster. Figure 6 shows an example of the information processing flow when performing faster responses within the premises shown in Figure 5. In the example shown in Figure 6, language data is provided from input cell 2-p to output cell 3-q. That is, the language data from input cell 2-p may be provided to output cell 3-q without going through the central AI61. Input cell 2-p and output cell 3-q are preferably equipped with the following basic logic, especially in an "autonomous decentralized AI blockchain cell."

[0080] Firstly, it is preferable that input cell 2-p and output cell 3-q have a "personality establishment logic". In other words, the "individuality establishment logic" is an algorithm for established logic processing that is individually provided for each input cell 2-p and output cell 3-q, according to each sensor's measurement target, etc. In other words, drive devices such as motors have variations in manufacturing and individual deterioration over time. Specifically, even among motors of the same model, some may vibrate and overheat abnormally at 1000 rpm (revolutions per minute), while others may rotate normally even at 2000 rpm. Each input cell 2-p, which generates various elements (physical quantities) through its own sensors attached to each motor, learns the individual characteristics of the motor it manages and can set its own limiter to determine what constitutes a normal operating range. When the input cell 2-p exceeds its individually set limiter, it can provide verbal data such as "motor rotation speed abnormal." Furthermore, the same individuality establishment logic can be applied to the stable operating temperature and other parameters of control boards other than motors.

[0081] Secondly, it is preferable that input cell 2-p and output cell 3-q have a "comparison learning logic". In other words, the "comparative learning logic" is an algorithm that understands the unique characteristics of the device it manages compared to other devices, based on the results of exchanging information with other cells. Specifically, input cell 2-p and output cell 3-q can understand the unique characteristics of the device they manage (the object measured by the sensor or the object controlled via the control device) compared to other devices by sharing data with other cells. This allows them to correct the output information based on the shared data. In other words, the language conversion AI model 75 of input cell 2-p performs a retraining process based on language data from other input cells, using the retraining execution unit 73. For example, using the example in Figure 1, suppose that when the language data "few people" is output, many other input cells, including input cell 2-2, are outputting the language data "many people". In this case, even though the area monitored by the sensor is the same, input cell 2-p is outputting the language data "few people". In such a case, the retraining execution unit 73 of input cell 2-p can perform a retraining process to make it easier to output "many people". In this way, input cell 2-p can learn based on (compared with) the language data output by other input cells (others).

[0082] Thirdly, it is preferable that input cell 2-p and output cell 3-q have a "consciousness / intention communication logic". The "Consciousness and Intention Communication Logic" is a logic algorithm in which the language conversion AI of input cell 2-p simultaneously outputs human language representing consciousness and intentions in addition to the normal digital data output from the sensor. Specifically, for example, the language conversion AI72 in input cell 2-p outputs language data such as "It's a little hot," "I feel a considerable vibration," and "This is a dangerous situation." This language data can be directly processed by the existing language input AI, the central AI61, allowing it to think and make judgments. In other words, in urgent situations, the edge server EDS in Figure 6 can notify the server 1 outside the premises (server 1 not shown in Figure 6, such as in Figure 4) and simultaneously perform corrective action without waiting for the higher-level decision result from the server 1 in the suburbs.

[0083] Fourth, it is preferable that input cell 2-p and output cell 3-q have "self-handling logic". This information processing system can take action without waiting for a response from the edge server EDS or server 1 (not shown) if a clear anomaly is detected, even if it does not require a decision from the central monitoring device. Specifically, for example, if input cell 2-p outputs language data such as "fire outbreak" instead of "temperature rise," that language data is provided to output cell 3-q, which can sound an alarm or activate local sprinklers. In other words, if we consider the central AI61 of this information processing system as the brain in humans, it becomes possible to respond in a way that could be described as a reflex, without waiting for judgment from the brain. This enables information processing for high-speed responses.

[0084] Fifth, it is preferable that input cell 2-p and output cell 3-q have "ethical logic". In other words, through ethical logic, the edge server EDS and the off-site server 1 (not shown) can perform total predictions and judgments, output output language data for control instructions, and provide training to each cell. Specifically, the language conversion AI model 75 of input cell 2-p is retrained by the retraining execution unit 73 based on the output language data from the central AI 61. Through this training, input cell 2-p becomes capable of outputting higher-level language data, such as "fire outbreak" instead of "temperature rise." As a result, input cell 2-p can make individual decisions and take action instantaneously according to the processing stage of input cell 2-p, thus preventing major accidents. In other words, this enables information processing for high-speed responses.

[0085] Thus, the input cell 2-p and output cell 3-q of this information processing system realize the concept of identity using IT, and can be called "autonomous decentralized AI blockchain cells."

[0086] Furthermore, the input cell 2-p and output cell 3-q of this information processing system function as "autonomous distributed AI blockchain cells," capable of autonomously determining processing content and requesting instructions for processing content from the central AI 61. Figure 7 shows an example of the process by which the input cell in the functional configuration of Figure 4 autonomously establishes the processing content.

[0087] As shown in step ST21, input cell 2-1 in Figure 7 acquires temperature information from the first sensor and humidity information from the second sensor. At this time, the autonomous distributed chip of input cell 2-1 autonomously recognizes its "mission" based on the information acquired from the sensor. Here, "mission" refers to the processing that input cell 2-1 should perform, specifically the processing that determines what kind of language data should be output from the temperature and humidity information. In other words, for example, input cell 2-1 identifies, based on the data format, that the acquired data is temperature and humidity information, and determines that its purpose is to determine the temperature (humidity) environment. Similarly, in input cell 2-2, once thermal information (such as thermographic data) is acquired from the sensor, input cell 2-2 determines that its mission is to determine the density of the heat source (specifically, the density of people in the example in Figure 1). In this way, input cells 2-1 and 2-2 can autonomously recognize their own mission.

[0088] As shown in step ST22, input cells 2-1 and 2-2 transmit their own mission to the central AI61.

[0089] As shown in step ST23, the central AI61 can integrate the missions transmitted from each cell to determine whether each cell's mission is valid. The central AI61 determines whether the mission autonomously recognized by each cell is valid or not. If the mission autonomously recognized by each cell is not valid, the central AI61 transmits information about the (correct) mission. To elaborate, the central AI61 can perform various decision-making processes in addition to the decision-making process for outputting output language data from input language data as described above. Specifically, for example, it can store information such as the overall layout of the equipment and the planned placement information of input cell 2-p, and perform processes such as integrating and deciding based on the information from the input cells.

[0090] As shown in step ST24, the cell's own mission is finalized by comparing the autonomously recognized mission with the (correct) mission information transmitted from the central AI61. Then, input cells 2-1 and 2-2 execute the subsequent processes based on the final determined mission.

[0091] In summary, input cells 2-1 and 2-2 can recognize their own mission based on information from the sensors. This establishes the individuality of input cells 2-1 and 2-2, allowing them to operate as cells dedicated to their respective sensors. Furthermore, when the sensor, conjunction, and information from the sensor are acquired, input cells 2-1 and 2-2 recognize their mission. Therefore, during production, input cells 2-1 and 2-2 can be shipped in a neutral state without the need for individual settings. This reduces production costs for input cells 2-1 and 2-2.

[0092] The embodiments of the "autonomous decentralized AI blockchain cell" have been described above, but the "autonomous decentralized AI blockchain cell" is not limited to the embodiments described above, and modifications, improvements, etc. may be made as appropriate to the extent that the objectives of the present invention can be achieved.

[0093] In the embodiment described above, input cell 2-p generates language data from sensor measurements, but it is not limited to this. That is, for example, the sensor may have a function to switch from a "language" for humans to a "language" for machines, and if an identifier is pre-associated in the information processing device, the sensor may convert the human language "temperature rise" to an identifier corresponding to "hot". This allows the information processing system to incorporate the results of checks performed by humans, etc., as input language data.

[0094] For example, the sensor may be a microphone or a camera. That is, for example, the language conversion AI 72 may identify a person from the voice acquired by the microphone and output it as language data, or identify the person's emotions and output them as language data. For example, the language conversion AI 72 may convert human gestures in a video acquired by the camera into language data.

[0095] As described above, an "autonomous decentralized AI blockchain cell" is a unit of information processing equipment that can perform processing related to sensors such as temperature, humidity, pressure, vibration, altitude, sound pressure, brightness, ultrasound, voltage, and current, as well as blockchain, based on the functions of its sensors. Although referred to as an information processing equipment, it is not limited to personal computers or server devices; it can be implemented as an electronic component chip. This allows for easy integration into various conventional devices.

[0096] Next, using Figures 8 to 11, we will describe an embodiment that includes an "autonomous decentralized AI blockchain audiovisual cell," that is, an input cell equipped with a camera. Furthermore, input cells equipped with a camera are specifically referred to as "input cell 2-r" (where r is an integer value between 1 and N, and not limited to p) to distinguish them from the aforementioned "input cell 2-p".

[0097] In other words, conventional unmanned monitoring of premises (factories, hospitals, apartment buildings, etc.), that is, monitoring without people being present on the premises, is used for security purposes and monitoring of various devices used on the premises. Specifically, captured images are recorded on a recorder, and human monitoring is performed to detect suspicious individuals, detect malfunctions, and predict malfunctions. Therefore, even in situations requiring instantaneous decisions, such as sudden incidents or accidents, human intervention can lead to delays in response, sometimes resulting in major disasters.

[0098] For example, traditionally, detecting suspicious individuals loitering around or individuals repeating dangerous behaviors has involved human personnel reviewing surveillance camera footage or checking recorded footage. Furthermore, in cases where suspicious individuals were loitering despite the area being off-limits to unauthorized personnel, or where someone who had previously engaged in dangerous behavior was repeating similar actions, the person in charge at the time, who had a grasp of the history of the situation, would make a judgment based on surveillance camera footage, or they would review the recordings after receiving a report of a suspicious person. Thus, the process of making a decision required the person in charge to review images from various surveillance cameras, which could have delayed the response and potentially increased the extent of the damage caused by the accident.

[0099] Automatic fire detection devices using heat sensors, smoke sensors, etc., have existed for some time. However, these conventional automatic fire detection devices only detect fires after they have occurred. In other words, conventional automatic fire detection devices cannot detect abnormalities before a fire occurs, and therefore cannot take preventative measures based on fire prediction. In other words, for example, in the case of an accident such as a fire, there are automatic detection devices that use heat sensors and smoke sensors, but while it is possible to detect anomalies after a fire has started, it is not possible to detect anomalies in advance, and it is impossible to take preventative measures based on prediction. Therefore, there was a need for a system that could initiate countermeasures more quickly and could be used without building a large, centralized system.

[0100] The "autonomous decentralized AI blockchain audiovisual cell" contributes to providing a system that more effectively processes image data captured by multiple cameras.

[0101] Embodiments of the present invention will be described below with reference to the drawings.

[0102] Figure 8 is a schematic diagram illustrating an example of the service to which the "autonomous decentralized AI blockchain audiovisual cell" is applied. This service, to which the "autonomous decentralized AI blockchain audiovisual cell" is applied, utilizes the information processing system shown in Figure 8 to provide a system that outputs the operation of a control device based on the input of image data captured by multiple cameras. In the explanation of Figure 8, the input cells equipped with cameras will be described using the same symbols as input cells 2-3 and 2-4 to distinguish them from the input cells described in Figure 1.

[0103] The information processing system for this service, as shown in Figure 8, consists of Server 1, input cells 2-3 and 2-4, output cell 3-1, verification terminal 4, and blockchain full nodes 5-1 to 5-4 (hereinafter referred to as "BC full nodes" as shown in Figure 1). Furthermore, as will be explained in more detail later, input cells 2-3 and 2-4, and output cell 3-1 are equipped with the functionality of a blockchain ultralight node (hereinafter referred to as "BC ultralight node" as shown in Figure 1).

[0104] Server 1 is an information processing device equipped with a central AI 61. As will be described in more detail later, Server 1 acquires data from images captured by cameras CAM-1 and CAM-2 in input cells 2-3 and 2-4, respectively, converted into perceptually represented data, as input data, and outputs perceptually represented control instructions in output cell 3-1 based on the input data as output data.

[0105] Input cells 2-3 and 2-4 are each equipped with cameras CAM-1 and CAM-2, and an autonomous distributed chip CP. The image data captured by the cameras is converted into perceptually represented data by the autonomous distributed chip CP and managed using the blockchain network BCN.

[0106] Output cell 3-1 is equipped with an autonomous distributed chip CP and an actuator (an example of a control device). The autonomous distributed chip CP receives control instructions output by server 1 from the blockchain network BCN and controls the control device (actuator) based on the control instructions.

[0107] Verification terminal 4 is an information processing device for verifying information managed using the blockchain network BCN. Equipment managers (such as those at factories) can use verification terminal 4 to verify the history of inputs and outputs that have not been tampered with.

[0108] Here, we will explain a specific example of "perceived data" in images captured by a camera. The image data (raw data) captured by multiple cameras consists of numerical color data for each pixel arranged in two dimensions in a still image. In contrast, humans can express the results of interpreting such images using language and other means. Specifically, for example, if an image of a certain man appears in an image of a first location for the first time, even though the image had never been included before that first time (for example, over a period of several months), a person viewing (watching) that image can perceive (recognize, interpret) and verbalize that a man who is not normally present at the first location is present at that first time. In this way, the perceived (recognized, interpreted) representation, rather than the image data (raw data), is called a "perceptual representation." Furthermore, the form of "language" is just one example of a form of perceptual representation. That is, for example, the data of the string "fire" is data that is perceptually represented in the form of the Japanese language. Also, for example, if identifiers are pre-assigned in an information processing device, the data of identifiers corresponding to "first location" or "fire" is also an example of perceptually represented data that can be used within the information processing device. Moreover, for example, the data of a still image of a flame icon can also be said to be data that is perceptually represented in the form of an icon corresponding to "fire".

[0109] First, the functions of the various nodes in the blockchain network BCN in this service, as shown in Figure 8, are basically the same as those in Figure 1. Furthermore, the functions of the BC ultralight nodes provided in the two input cells 2-3 and 2-4, and the output cell 3-1, respectively, in the example shown in Figure 8 are basically the same. Therefore, a detailed explanation is omitted.

[0110] In summary, Figure 8 illustrates an example of the configuration of the information processing system in this service to which the "autonomous distributed AI blockchain cell audiovisual cell" is applied. The following details of the information processing flow in this service will be explained using Figure 8, following steps ST21 to ST29.

[0111] In step ST21, it is assumed that camera CAM-1, located in input cell 2-3, is capturing an image of the first location at the first time step. The image data captured by camera CAM-1 is then obtained as data for "an image containing the image of a certain man".

[0112] In step ST22, the language conversion AI of the autonomous distributed chip CP in input cells 2-3 converts the image data (digital signal) into language. For example, suppose the image of a man contained in the image had never been captured before the first time step. In such a case, the image data (digital signal) is converted into language data that says, "At the first time step, there is a man at the first location who is not normally there."

[0113] In step ST23, the BC ultralight node of the autonomous distributed chip CP in input cell 2-3 manages the linguistic data "There is a man at location 1 at time 1 who is not normally there" using the blockchain network BCN. Specifically, for example, it manages the linguistic data "There is a man at location 1 at time 1 who is not normally there" by sending it to BC full node 5-1 as part of a transaction in blockchain technology. At this time, the BC ultralight node of the autonomous distributed chip CP in input cell 2-3 manages the encrypted linguistic data using the blockchain network BCN. Furthermore, at this time, the characteristics of the men (height, build, clothing characteristics, etc.) may be managed as appropriate. Furthermore, the image data itself may be managed using the blockchain network BCN as appropriate. Specifically, for example, the metadata of the image data may be managed using the blockchain network BCN. The image data itself may then be encrypted or partitioned as appropriate and managed using IPFS or similar.

[0114] Although not shown in the diagram, the same processing as in steps ST21 to ST23 described above is also performed in input cells 2-4. Specifically, for example, suppose camera CAM-2, located in input cell 2-4, captures an image of the second location at a second time point, which is later than the first time point. Then, the image data captured by camera CAM-2 includes data of "an image containing the image of a certain man." Then, the language conversion AI in the autonomous distributed chip CP of input cells 2-4 converts the image data (digital signal) into language. For example, suppose the image of a man contained in the image had never been captured before the first time step. In such a case, the image data (digital signal) is converted into language data that says, "At the second time step, there is a man at the second location who is not normally there." Then, the BC ultralight nodes on the autonomous distributed chip CP in input cells 2-4 manage the encrypted linguistic data "There is a man at location 2 at time 2 who is not normally there" using the blockchain network BCN.

[0115] In step ST24, Server 1 decrypts and retrieves encrypted language data managed using the blockchain network BCN. In step ST25, Server 1 uses the blockchain network BCN to manage the results of its determination of the content of control instructions based on the acquired language data. Specifically, for example, Server 1 uses Central AI 61, which has been trained as an AI that performs natural language processing, to generate linguistic data of a control instruction, "Warn the man at location 2," as output data, using the linguistic data "At time 1, there is a man at location 1 who is not normally there" and "At time 2, there is a man at location 2 who is not normally there." Server 1 then manages the generated control instruction, "Warn the man at location 2," using the blockchain network BCN. At this time, Server 1 manages the encrypted control instruction using the blockchain network BCN.

[0116] In step ST26, the BC ultralight node of the autonomous distributed chip CP in output cell 3-1 obtains encrypted control instructions managed using the blockchain network BCN.

[0117] In step ST27, the firmware of the autonomous distributed chip CP in output cell 3-1 decrypts the encrypted control instructions. Then, the language conversion AI of the autonomous distributed chip CP in output cell 3-1 converts them into digital signals as specific control content for the control device. Specifically, for example, the language conversion AI in the autonomous distributed chip CP of output cell 3-1 converts a control instruction to "warn the man at the second location" into a digital signal that causes a control device (such as an actuator or the control unit of an alarm device) to operate to "activate an alarm device" around the second location. In other words, for example, the language conversion AI in the autonomous distributed chip CP of output cell 3-1 outputs a digital data control signal (digital signal) to activate the alarm device. Although not shown in the diagram, the BC ultralight node in the autonomous distributed chip CP of output cell 3-1 can also manage the operation of the control device, which "activates alarm devices around the second location," using the blockchain network BCN.

[0118] In step ST28, the actuator (an example of a control device) is driven in response to a digital data control signal (digital signal) that causes the alarm device around the second location to activate. As a result, the actuator (an example of a control device) operates in response to the control instruction from the central AI61 to "warn the man at the second location." Based on the image data captured by the camera, such as "a man who is not normally present is at location 1 at time 1" and "a man who is not normally present is at location 2 at time 2," an alarm system around location 2 (not shown) is activated to warn the man.

[0119] In step ST29, the verification terminal 4 presents the information managed using the blockchain network BCN to the facility (factory, etc.) manager. Specifically, the facility (factory, etc.) manager can verify the following information presented to the verification terminal 4: image data from the cameras at the first and second locations for the first and second time points, input data (linguistic data) such as "There is a man at the first location at the first time point who is not normally there" and "There is a man at the second location at the second time point who is not normally there," output data (linguistic data) for the control instruction "Warn the man at the second location," and the operation of the control device that "Activates the alarm device" around the second location. Because this information is managed using the blockchain network BCN, it is presented to the facility (factory, etc.) manager as an untampered record of input and output.

[0120] With the configuration and operation described above, this service has the following characteristics, as explained in the description of Figure 1 above. Firstly, an existing AI that has been trained in language can be adopted as the central AI 61. Therefore, this service can reduce overall development and operational costs.

[0121] Secondly, because AI processing is performed on the edge side, the processing burden on the central AI61 is reduced. In other words, the central AI61 does not need to process perceptual representation (verbalization), thus reducing its workload. This means that distributed processing of AI is being achieved. Furthermore, the effect of this distributed processing becomes more pronounced as the number of input and output cells on the edge side increases.

[0122] Thirdly, this service uses the blockchain network BCN to guarantee data integrity, thus realizing a secure AI system. As a result, this service uses the blockchain network BCN to guarantee data, making it impossible to impersonate input cells 2-3, thus realizing a secure AI system.

[0123] In the example above, the language AI was used to explain that it produced an output indicating that a man who is not normally present was at a certain location at a certain time. However, language AI can produce a wide variety of outputs. Specifically, it may output perceptual expressions such as, "Sparks are flying (from the installed device) at location 1," "Smoke is being emitted at location 1," or "A fire has broken out at location 1."

[0124] The above explains the overview of this service to which the "autonomous decentralized AI blockchain cell" is applied, using Figure 8. The configuration of the information processing system applied when providing the service shown in Figure 8 of this service, to which the "autonomous decentralized AI blockchain cell" is applied, is basically the same as that shown in Figures 2 and 3.

[0125] However, as shown in Figure 1, input cell 2-r is equipped with camera CAM-p as an input unit. Furthermore, input cell 2-r has some or all of the CPU, ROM, RAM, etc., as an autonomous distributed chip CP.

[0126] Through the collaboration of various hardware and software components of Server 1, various processes can be executed. As a result, the aforementioned service can be provided. The functional configuration of the information processing system providing the service shown in Figure 8 will be described below.

[0127] Figure 9 is a functional block diagram showing an example of the functional configuration of the information processing system that provides the service shown in Figure 8.

[0128] As shown in Figure 9, the CPU 11 of server 1 has a processing execution unit 51 that functions. The memory unit 18 stores the model of the central AI 61.

[0129] In input cell 2-r, the autonomous distributed chip CP functions as a model management unit 71, a language conversion AI 72, and a retraining execution unit 73. Input cell 2-r has a camera CAM-p as its input unit. The language conversion AI model 75 is stored in the memory unit of input cell 2-r.

[0130] In output cell 3-q, the autonomous distributed chip CP functions as a model management unit 81, a reverse language conversion AI 82, a control unit 83, and a retraining execution unit 84. Output cell 3-q has an actuator 85 as an output unit. In addition, the language conversion AI model 86 is stored in the memory unit of output cell 3-q.

[0131] The processing execution unit 51 acquires input language data from one or more input cells 2-r as input data, executes a predetermined process using the input data, and outputs one or more output language data indicating the execution result of that process as output data. The processing execution unit 51 retrieves the input data stored using the blockchain network BCN.

[0132] The model management unit 71 takes image data captured by the camera CAM-p as input, converts it into language data, and stores and manages the language conversion AI model 75 in the memory unit. Specifically, for example, if the model management unit 71 includes an image of a man that has never been captured before, it converts it into linguistic data such as "There is a man at location 1 at time 1 who is not normally there" and outputs it.

[0133] The language conversion AI 72 acquires image data of the target region output from the camera CAM-p that captures the target region, inputs it into the language conversion AI model 75, and outputs the language data output from the model as at least part of the input language data of the server 1. The processing after the language conversion by AI72 is basically the same. Therefore, the explanation will be omitted. The following describes an example of applying this information processing system to monitor multiple devices located within a premises, using Figures 10 and 11. Figure 10 shows an example of using the service shown in Figure 1 to manage equipment located within the premises. In the example shown in Figure 10, two devices, A1 and A2, are located within the premises. That is, for example, devices A1 and A2 are each piece of equipment on a manufacturing line located within the factory premises.

[0134] Furthermore, an input cell 2-1 is located on device A1. This shows, for example, the camera CAM-1 of input cell 2-1 capturing images of device A1 on the manufacturing line and its surroundings as the target area. Specifically, for example, the sensor captures images of device A1, capturing its appearance, a predetermined meter, and the conditions around device A1.

[0135] Furthermore, device A2 is equipped with input cells 2-2 and 2-3. This shows that device A2 is being imaged by two cameras, camera CAM-2 and infrared camera CAM-3, which are connected to the two input cells 2-2 and 2-3. In other words, while device A1 was equipped with one input cell 2-1, device A2 is equipped with multiple input cells. In this case, input cell 2-5 is imaging the target area with the infrared camera CAM-3. As a result, input cell 2-5 can acquire images that allow for the detection of events such as abnormal heat generation occurring inside device A2.

[0136] Then, each input cell 2-1 to 2-3 converts the data measured by the sensor into language data and transmits it to the edge server EDS. In this way, data from multiple devices A1 and A2 within the premises is collected.

[0137] This enables fault detection and prediction based on the operating status of various devices used within the premises (e.g., manufacturing robots (lines), room temperature control, lighting control, etc.). Furthermore, it enables unmanned monitoring of the premises (e.g., factories, hospitals, apartment buildings, etc.) as shown in Figure 5. As mentioned above, in conventional monitoring systems, when an accident occurs and data (abnormal values) is transmitted simultaneously from many sensors, the central monitoring device processes the data sequentially, which can result in a considerable delay before it can issue an alarm to the user (e.g., a monitoring officer) or activate emergency equipment (e.g., fire suppression using sprinklers).

[0138] In contrast, this information processing system distributes processing because each input cell is converted into language data, enabling high-speed processing overall.

[0139] Furthermore, by processing without going through control instructions from the central AI61, it is possible to handle the situation even faster. Figure 11 shows an example of the information processing flow when performing faster responses within the premises shown in Figure 10. In the example shown in Figure 11, language data is provided from input cell 2-r to output cell 3-q. That is, the language data from input cell 2-r may be provided to output cell 3-q without going through the central AI61. Input cell 2-r and output cell 3-q, in particular, in an "autonomous decentralized AI blockchain audiovisual cell," preferably have the following basic logic.

[0140] Firstly, it is preferable that input cell 2-r has an "autonomous anomaly detection logic". In other words, the "autonomous anomaly detection logic" is a logic that autonomously analyzes the camera image for each input cell 2-r and processes an alarm. As mentioned above, input cell 2-r can detect an abnormal (unusual) state if the image contains a man who is not normally present. For example, input cell 2-r could also detect overheating of equipment, fire, abnormal behavior (such as acting violently, crying for help, or being chased), intrusion into restricted areas, or intrusion of dangerous animals. Furthermore, as described above using Figure 5, by using an infrared camera in addition to a conventional camera capable of capturing visible light as shown in Figure 1, it becomes possible to detect phenomena that cannot be detected by conventional cameras, such as abnormal overheating inside and outside the device, or high-temperature fires that are invisible on the other side of a wall. Furthermore, it is preferable that input cell 2-r also has a microphone as an input unit. This allows for the use of, for example, human voices or sounds generated by accidents, in conjunction with images, thereby improving the accuracy of detection.

[0141] Secondly, it is preferable that input cell 2-r has a "past comparison learning (self-learning) logic". In other words, the "past comparison learning (self-learning) logic" is a logic that activates the relearning execution unit 73 of input cell 2-r and learns to perform detection based on information previously obtained from input cell 2-r itself and other input cells. Specifically, for example, the retraining execution unit 73 receives an image of a predetermined target person and information about the characteristics of that person from a higher-level server (for example, the edge server EDS in Figure 5), and performs retraining (self-learning) to detect whether the image of that target person is actually included in the image. In addition, even if the input cell 2-r detects during retraining (self-learning), it can also activate the output cell 3-q to sound an alarm, as shown in Figure 6.

[0142] Thirdly, it is preferable that this system has an "automatic tracking logic". The automatic tracking logic is a logic that automatically tracks a predetermined target in each of the multiple input cells 2-r, using information indicating that a predetermined target has been detected in each input cell. Specifically, for example, a higher-level server (e.g., the edge server EDS in Figure 5) collects information from each of the multiple input cells 2-r that has detected a target person or other object. The higher-level server then automatically tracks the target person by estimating their movement path and future location based on the information that the target person or other object has been detected at each predetermined location in the area imaged by each input cell 2-r.

[0143] Fourth, it is preferable that this system has a "consciousness and communication logic". The consciousness / intention communication logic is a logic that simultaneously outputs digital image data as well as human language representing consciousness and intentions in each of the multiple input cells 2-r. Specifically, as shown in Figure 1, input cell 2-r manages the captured image data (including still and moving images) using the blockchain network BCN, and simultaneously outputs linguistic expressions corresponding to human consciousness and intentions. Specifically, for example, linguistic expressions such as "Abnormal temperature detected" or "Person moving violently detected" are output from input cell 2-r. By outputting in this linguistic form, as mentioned above, it becomes possible to connect it directly to an existing central AI capable of language input and enable it to think and make decisions. Here, we will explain the difference between the edge server EDS and the central AI 61 in Figures 5 and 6. That is, as mentioned above, the central AI 61 usually makes the basic decisions. However, the edge server EDS can operate output cell 3-q via the edge server EDS when it is not necessary to wait for a decision from the central AI 61.

[0144] Fifth, it is preferable that this system has a "self-handling logic". Self-handling logic is logic that executes actions in each of the multiple input cells 3-q without going through the central AI61 or the edge server EDS. Specifically, unlike ordinary surveillance cameras, output cell 3-q, as explained using Figure 6, can issue an alarm or perform localized firefighting using sprinklers, etc., if there is a dangerous situation that requires immediate action without needing to consult the central AI61. In other words, this enables information processing for high-speed responses.

[0145] Thus, the input cell 2-r and output cell 3-q of this information processing system are equipped with audiovisual features such as cameras and microphones, realizing the concept of identity through IT, and can be called "autonomous decentralized AI blockchain audiovisual cells."

[0146] Although one embodiment of the present invention has been described above, the present invention is not limited to the embodiments described above, and any modifications, improvements, etc. that can achieve the objectives of the present invention are considered to be included in the present invention.

[0147] In the embodiment described above, input cell 2-r generates language data from image data captured by the camera, but it is not limited to this. That is, for example, input cell 2-r may have a function to convert from a "language" for humans to a "language" for machines. Specifically, for example, input cell 2-r may obtain from a human that there is a "suspicious person" in human language and convert it to an identifier that indicates there is a "suspicious person" which is pre-associated in the information processing device. This allows the information processing system to incorporate the results of checks performed by humans, etc., as input language data.

[0148] As described above, an "autonomous decentralized AI blockchain audiovisual cell" is a unit of information processing equipment capable of executing blockchain-related processing. While referred to as an information processing equipment, it is not limited to personal computers or server devices; it can be implemented as an electronic component, such as a chip. This allows for easy integration into various existing devices. In other words, the autonomous distributed chip CP in Figure 1 may be implemented as a chip and operate as an input cell when embedded in a camera.

[0149] Such "autonomous decentralized AI blockchain cells" and "autonomous decentralized AI blockchain audiovisual cells" can be connected to each other within the premises via a predetermined network, particularly a wireless network, thereby enabling autonomous decentralized processing and significantly improving convenience.

[0150] Regardless of whether a wired connection, a wireless connection, or a combination thereof is used for network connectivity, in order to manage, integrate, and effectively utilize a large number of input cells 2 ("autonomous decentralized AI blockchain cells" and "autonomous decentralized AI blockchain audiovisual cells"), it is necessary for the system to perform functions that facilitate collaboration.

[0151] In other words, in conventional systems where a centralized server manages a large number of sensors, the load on the centralized server increases as the number of sensors increases. However, as described above, by using input cell 2 of this embodiment, the benefits of distributed AI become apparent as the number of sensors increases, and the load on the centralized server (central AI in this embodiment) decreases. This is the greatest advantage of input cell 2 of this embodiment. In other words, the most advantageous point of the input cell 2 in this embodiment is that it allows for the easy construction of a decentralized system rather than a conventional centralized system. Input cell 2 and output cell 3-q, which include the aforementioned "autonomous decentralized AI blockchain cell" and "autonomous decentralized AI blockchain audiovisual cell," preferably have the following basic logic.

[0152] Firstly, the information processing system of this embodiment preferably includes a "self-position recognition logic". In other words, for input cell 2 to autonomously detect anomalies, determine solutions, and take action, it is crucial for input cell 2 to understand "where it is" on its own. Specifically, for example, if a fire breaks out on the third floor, the priority processing will naturally differ between input cell 2-1 on the third floor and input cell 2-2 on the first floor. In other words, distributed processing tailored to each location and role is necessary.

[0153] Therefore, it is preferable for input cell 2 to be able to understand the relationship between the overall map and its own position, such as which device or location it is installed in on the map, and which other input cells 2 are nearby, through the overall map. This enables autonomous control (for example, control processing from output cell 3) by appropriately utilizing the information of multiple input cells 2. In performing this overall map mapping, it is preferable that each input cell 2 be equipped with means for acquiring information to determine its 3D position, such as GPS or altitude sensors. This allows input cell 2 itself to identify and understand its own location (position) within the 3D map, as well as the location and device where anomalies are detected, not only through notifications from higher-level servers (e.g., edge server EDS or server 1). The map should be created, updated, and managed by a higher-level server. By periodically notifying each "autonomous decentralized AI blockchain cell," the entire system can use a unified map and perform distributed processing according to its specific needs. Furthermore, by performing this 3D mapping, it becomes possible to share positional information between input cells 2 attached to mobile objects such as flying drones. This makes it possible to utilize input cells 2 to realize safe operation and proxy operation.

[0154] Secondly, the information processing system of this embodiment preferably includes an "input cell education logic". In other words, it is preferable for the higher-level servers located inside or outside the premises that manage input cells 2 (e.g., edge server EDS or server 1) to periodically notify all input cells 2 of the latest map, which is used to comprehensively map the locations of each facility or department (rooms and installation locations in the case of an apartment building) in 3D based on 3D coordinate information such as position and altitude information from each input cell, and to share this map with all input cells 2. In order to keep the map up-to-date at all times, the 3D map can be updated based on information such as whether it is functioning correctly, whether its location has changed, and its status, by periodically polling input cell 2. This means that, for example, if a faulty input cell 2 is replaced or a new input cell 2 is added, the newly installed input cell 2 can autonomously understand its own position and role and operate without human intervention.

[0155] Thirdly, the information processing system of this embodiment preferably has a "distributed autonomous system". Here, a distributed autonomous system can be called a "Decentralized Autonomous System (DAS)," and specifically it is as follows: In other words, as described above, input cell 2 has logic that allows it to make autonomous (independent) decisions and perform processing. Therefore, even if a local server (e.g., an edge server EDS) temporarily becomes inoperable due to a failure or maintenance, input cell 2 or output cell 3 can monitor and respond individually, or in cooperation with each other. Furthermore, when the local server (e.g., edge server EDS) recovers (normally resumes operation), input cell 2 and output cell 3 can instantly restore the latest mapping and state from information distributed and maintained on the higher-level server (e.g., edge server EDS), enabling them to return to normal operation. The greatest benefit of input cell 2 and output cell 3, and arguably the greatest advantage of a distributed autonomous system (DAS), is that they enable the system to carry out its mission without failure, even without a higher-level server (e.g., an edge server EDS).

[0156] Fourth, it is preferable that the information processing system of this embodiment has an "ethical logic". In other words, higher-level servers both inside and outside the campus (for example, edge server EDS and server 1) perform overall predictions and judgments, and then, based on their own analysis and predictions, provide the most suitable education to each input cell 2 and output cell 3. The reason why education in this context refers to the fact that the things that need to be recognized vary depending on the device and environment in which each input cell 2 and output cell 3 are installed. Specifically, even if the sensor information is the same 30 degrees, its meaning can differ depending on the installation location, such as whether the input cell 2 is installed indoors or outdoors. Furthermore, terms like noise and vibration can have different meanings depending on the environment and the machine. Therefore, the perceptual representation output by input cell 2 should be corrected to produce language output that is as close as possible to the tactile sensations a human would perceive in that moment. Furthermore, since there are specific terms and vocabulary unique to each industry and type of equipment, it is a good idea to educate employees while correcting them for these specific expressions. Ideally, input cell 2 and output cell 3 should be able to receive correction information from these higher-level servers (e.g., edge server EDS or server 1) and self-learn to produce output language that is aware of their own location, environment, and equipment.

[0157] Furthermore, the system configuration shown in Figure 2 and the hardware configuration of Server 1 shown in Figure 3 are merely illustrative examples for achieving the objectives of the present invention and are not particularly limited.

[0158] Furthermore, the functional block diagrams shown in Figures 4 and 9 are merely illustrative and not particularly limiting. In other words, it is sufficient that the information processing system in Figure 2 has the functionality to execute the series of processes described above as a whole, and the functional blocks and databases used to realize this functionality are not particularly limited to the examples in Figures 4 and 9.

[0159] Furthermore, the location of the functional block is not limited to Figure 5, but can be any location. For example, at least a portion of the functional blocks located on Server 1 may be provided by other information processing devices.

[0160] Furthermore, the series of processes described above can be executed by hardware or by software. Furthermore, a single functional block may consist of hardware alone, software alone, or a combination of both.

[0161] When a series of processes are executed by software, the programs that make up that software are installed on a computer or other device from a network or storage medium. The computer may be a computer that is built into dedicated hardware. Furthermore, a computer can be any computer capable of performing various functions by installing various programs, such as a server, a general-purpose smartphone, or a personal computer.

[0162] Such recording media containing programs may consist not only of removable media (not shown) distributed separately from the main unit of the device to provide the program to the user, but also of recording media provided to the user in a state where they are pre-installed in the main unit of the device.

[0163] In this specification, the step of describing a program to be recorded on a recording medium includes not only processes that are performed chronologically in that order, but also processes that are not necessarily performed chronologically, but are executed in parallel or individually.

[0164] In summary, the information processing system to which the present invention applies only needs to have the following configuration, and can take various forms.

[0165] In other words, the information processing system to which the present invention is applied (for example, the information processing system in Figures 4 and 9) is: In an information processing system including a central device that performs predetermined processing using input data (for example, Server 1 in Figures 4 and 9, or Edge Server EDS in Figures 6 and 10), and one or more Type 1 peripheral devices that provide at least a portion of the input data to the central device (for example, Input Cell 2 in Figures 4 and 9), The aforementioned central device is A processing execution means (processing execution unit 51) acquires one or more perceptual representation data (input language data in Figures 4 and 9) as input data, performs a predetermined process using the input data, and outputs one or more perceptual representation data indicating the result of the process as output data. Equipped with, Each of the one or more Type 1 peripheral devices is: A model management means (model management unit 71) that takes predetermined data (for example, temperature data or image data) as input, converts it into the aforementioned perceptual representation data (for example, data corresponding to "temperature rise" or data corresponding to "suspicious person present"), and outputs a model (for example, a language conversion AI model 86) is stored and managed in a predetermined storage medium. A conversion means (language conversion AI 72) acquires data (e.g., temperature data or image data) output from a sensor that measures physical quantities in the real world (e.g., sensor 74 in Figure 4) or a camera that images a target area (e.g., camera in Figure 9) and inputs it into the model, and outputs the perceptual representation data (e.g., data corresponding to "temperature rising" or "suspicious person present") output from the model as at least part of the input data of the central device, Having that will suffice. This improves the convenience of managing equipment that is managed based on data from multiple sensors or cameras.

[0166] The information processing system further includes one or more Type 2 peripheral devices (output cells 3) that control a predetermined control target, Each of the above one or more Type 2 peripheral devices is: A model management means (model management unit 81) that takes the aforementioned perceptual representation data as input, converts it into a predetermined physical quantity, and outputs a model, which is then stored and managed in a predetermined storage medium. A control means (reverse language conversion AI 82 and control unit 83) controls a predetermined control target by acquiring at least a portion of the one or more perceptual representation data that constitute the output data output from the central device and inputting it into the model, and inputting the signal indicating the predetermined physical quantity output from the model as an instruction signal to the predetermined control target, It can be equipped with. This further improves the convenience of managing equipment that is managed based on data from multiple sensors or cameras.

[0167] The aforementioned central device is A first peripheral device information acquisition means (e.g., a processing execution unit 51) acquires position information (e.g., 3D position information obtainable from GPS or an altitude sensor) and information relating to the model (e.g., the model itself, which is data that affects the threshold for determining temperature rise) from one or more first peripheral devices as first peripheral device information, A map management means (e.g., a processing execution unit 51) that generates, updates, and manages a three-dimensional map of the facility on which the one or more Type 1 peripheral devices are arranged, based on the Type 1 peripheral device information, Furthermore, Each of the one or more Type 1 peripheral devices mentioned above is: A location information acquisition means (e.g., model management unit 71) that acquires location information of the first type peripheral device, A first type peripheral device information transmission means (e.g., a model management unit 71) that provides the aforementioned location information and information regarding the aforementioned model to the central device, The central device manages a 3D map, and a first type of retraining execution means (for example, the retraining execution unit 73 in Figure 4) performs retraining to update the model based on the position information. It can provide even more. This further improves the convenience of managing equipment that is managed based on data from multiple sensors or cameras. [Explanation of symbols]

[0168] 11...Server, 2...Input cell, 3...Output cell, 4...Verification terminal, 5...Blockchain full node, 11...CPU, 20...Drive, 31...Removable media, 51...Processing execution unit, 61...Central AI, 71...Model management unit, 72...Language conversion AI, 73...Retraining execution unit, 74...Sensor, 75...Language conversion AI model, CAM...Camera, 81...Model management unit, 82...Reverse language conversion AI, 83...Control unit, 84...Retraining execution unit, 85...Actuator, 86...Reverse language conversion AI model

Claims

1. An information processing system comprising a central device that performs predetermined processing using input data, one or more first-type peripheral devices that provide at least a portion of the input data to the central device, and one or more second-type peripheral devices that control a predetermined control target, The aforementioned central device is A processing execution means that acquires multiple perceptual representation data as input data, performs a process to integrate or analyze the multiple perceptual representation data, and outputs one or more perceptual representation data representing the result of the execution of that process as output data, A means for acquiring type 1 peripheral device information that acquires position information and parameter information of the first model from one or more of the aforementioned type 1 peripheral devices as type 1 peripheral device information, A three-dimensional map of a facility on which one or more Type 1 peripheral devices are arranged, comprising a map management means for generating, updating, and managing a three-dimensional map showing the arrangement locations of the Type 1 peripheral devices based on the location information, A relearning data transmission control means that performs control to transmit to a predetermined type 1 peripheral device, as relearning data, at least one of the installation environment and role of a predetermined type 1 peripheral device identified based on the three-dimensional map, from the parameter information obtained from each of the one or more type 1 peripheral devices, the perceptual representation data obtained as input data from each of the one or more type 1 peripheral devices, and the perceptual representation data as output data showing the result of the processing using the input data, Equipped with, Each of the one or more Type 1 peripheral devices mentioned above is: A model management means that takes predetermined data as input, converts it into the perceptual representation data, and outputs the first model, which is then stored and managed in a storage medium within the first type peripheral device. A conversion means that acquires data output from a sensor that measures physical quantities in the real world, or from a camera that images a target area, inputs it into the first model, and outputs the perceptual representation data output from the first model as at least a part of the input data of the central device, A location information acquisition means for acquiring the location information of the first type peripheral device, A first type peripheral device information transmission means that provides the aforementioned location information and the aforementioned parameter information of the first model to the central device, A first type of retraining execution means that receives the retraining data from the central device and performs retraining to update the first model, Equipped with, Each of the one or more Type 2 peripheral devices mentioned above is: A model management means that takes the aforementioned perceptual representation data as input, converts it into a predetermined physical quantity, and outputs a second model, which is then stored and managed in a predetermined storage medium. A control means for controlling a predetermined control target by acquiring at least a portion of the one or more perceptual representation data constituting the output data output from the central device and inputting it into the second model, and inputting the signal indicating the predetermined physical quantity output from the second model as an instruction signal to the predetermined control target, Equipped with, Information processing system.

2. An information processing method performed by an information processing system including a central device that performs predetermined processing using input data, one or more first-type peripheral devices that provide at least a portion of the input data to the central device, and one or more second-type peripheral devices that control a predetermined control target, The steps performed by the central device include: A processing execution step which involves acquiring multiple perceptual representation data as input data, performing a process to integrate or analyze the multiple perceptual representation data, and outputting one or more perceptual representation data representing the result of the process as output data, A first peripheral device information acquisition step is performed to acquire position information and parameter information of the first model from one or more first peripheral devices as first peripheral device information, A map management step of generating or updating a three-dimensional map of a facility on which one or more Type 1 peripheral devices are arranged, which shows the arrangement locations of the Type 1 peripheral devices based on the location information, A relearning data transmission control step that executes control to transmit to a predetermined type first peripheral device, as relearning data, at least one of the installation environment and role of a predetermined type first peripheral device identified based on the three-dimensional map, from the parameter information obtained from each of the one or more type first peripheral devices, the perceptual representation data obtained as input data from each of the one or more type first peripheral devices, and the perceptual representation data as output data showing the result of the processing using the input data, Includes, As a step performed by each of the one or more Type 1 peripheral devices, A model management step involves inputting predetermined data, converting it into the aforementioned perceptual representation data, and outputting a first model, which is then stored and managed in a storage medium within the first type peripheral device. A conversion step of acquiring data output from a sensor that measures physical quantities in the real world, or from a camera that images a target area, inputting it into the first model, and outputting the perceptual representation data output from the first model as at least a part of the input data of the central device, A position information acquisition step for acquiring the position information of the first type peripheral device, A first type peripheral device information transmission step provides the location information and the parameter information of the first model to the central device, A first type of retraining execution step in which the retraining data is received from the central device and retraining is performed to update the first model, Includes, As a step performed by each of the one or more Type 2 peripheral devices mentioned above, A model management step involves inputting the aforementioned perceptual representation data, converting it into a predetermined physical quantity, and outputting a second model, which is then stored and managed in a predetermined storage medium. A control step which involves acquiring at least a portion of the one or more perceptual representation data constituting the output data output from the central device and inputting it into the second model, and inputting the signal indicating the predetermined physical quantity output from the second model as an instruction signal to the predetermined control target in order to control the predetermined control target, including, Information processing methods.

3. An information processing system comprising a central device that performs predetermined processing using input data, one or more Type 1 peripheral devices that provide at least a portion of the input data to the central device, and one or more Type 2 peripheral devices that control predetermined controlled objects, includes a computer. The processes to be performed by the aforementioned central device are as follows: A process execution process that takes multiple perceptual representation data as input data, performs a process to integrate or analyze the multiple perceptual representation data, and outputs one or more perceptual representation data representing the result of the execution of that process as output data, A first-type peripheral device information acquisition process that acquires position information and parameter information of the first model from one or more of the first-type peripheral devices as first-type peripheral device information, A map management process that generates or updates a three-dimensional map of a facility on which one or more Type 1 peripheral devices are arranged, and which manages the arrangement of the Type 1 peripheral devices based on the location information, A retraining data transmission control process that executes control to transmit to a predetermined type 1 peripheral device, as retraining data, at least one of the installation environment and role of a predetermined type 1 peripheral device identified based on the 3D map, from the parameter information obtained from each of the one or more type 1 peripheral devices, the perceptual representation data obtained as input data from each of the one or more type 1 peripheral devices, and the perceptual representation data as output data showing the result of the processing using the input data, Make it run, The processing performed by each of the one or more Type 1 peripheral devices is as follows: A model management process that takes predetermined data as input, converts it into the aforementioned perceptual representation data, and outputs a first model, which is then stored and managed in a storage medium within the first type peripheral device. A conversion process that acquires data output from a sensor that measures physical quantities in the real world, or from a camera that images a target area, inputs it into the first model, and outputs the perceptual representation data output from the first model as at least a part of the input data of the central device, A position information acquisition process for acquiring the position information of the said Type 1 peripheral device, A first type peripheral device information transmission process that provides the aforementioned location information and the parameter information of the first model to the central device, A first type of retraining execution process that receives the retraining data from the central device and performs retraining to update the first model, Make it run, The processing performed by each of the one or more Type 2 peripheral devices is as follows: A model management process that takes the aforementioned perceptual representation data as input, converts it into a predetermined physical quantity, and outputs a second model, which is then stored and managed in a predetermined storage medium. A control process is performed to control a predetermined control target by acquiring at least a portion of the one or more perceptual representation data that constitute the output data output from the central device and inputting it into the second model, and inputting the signal indicating the predetermined physical quantity output from the second model as an instruction signal into the predetermined control target, To execute program.

Images